An Adaptive Control Method for an Aircraft and a Computing Device

Abstract

The embodiments of this specification provide an adaptive control method for an aircraft and a computing device, which determine the real aircraft state and the simulated aircraft state corresponding to the current control period. The real aircraft state is determined by an on-board sensor, and the simulated aircraft state is determined by a reference model. An adaptive controller is used to determine the control target compensation amount corresponding to the current control period according to the difference between the real aircraft state and the simulated aircraft state. According to the control target compensation amount and the initial control target corresponding to the current control period, the corrected control target corresponding to the current control period is determined. An inner-loop controller is used to control the aircraft according to the corrected control target and the real aircraft state, and the reference model is used to simulate the aircraft state of the next control period according to the corrected control target and the simulated aircraft state. It can provide accurate control commands under the conditions of large flight envelopes and strong couplings, and improve the anti-interference ability of the aircraft.

Description

Technical Field

[0001] The embodiments of this specification belong to the technical field of data processing, and more specifically, relate to an adaptive control method for an aircraft and a computing device. Background Art

[0002] Electric Vertical Takeoff and Landing (eVTOL) aircraft are increasingly widely used in urban air transportation and the drone field, and the requirements for the robustness and anti-interference ability of their flight control systems are also increasing day by day. Currently, developers usually use the Model Reference Adaptive Control (MRAC) method to handle the dynamic characteristics and parameter uncertainties of aircraft in complex environments to improve the robustness of the flight control system.

[0003] However, traditional model reference adaptive flight control systems are generally designed based on linear models or low-order transfer functions. During the design process, it is necessary to trim and linearize the non-linear flight dynamics model near the operating point (a specific aircraft state, such as level flight or climbing at a fixed angle), and then configure the aircraft adaptive control law according to the linearized model obtained after the linearization process. It can be seen that the linearized model is a simplified model obtained by linearizing the original high-precision non-linear flight dynamics model based on the "small perturbation" assumption. Therefore, the higher the non-linearity of the original flight dynamics model, the more obvious the error between the linear model and the non-linear model, and the less matching the calculated adaptive control law is with the actual response of the aircraft. The control accuracy and anti-interference ability of traditional model reference adaptive flight control systems are difficult to meet the requirements under large envelope or strong coupling working conditions.

[0004] Therefore, the present invention provides an adaptive control method for an aircraft and a computing device. Summary of the Invention

[0005] The embodiments of this specification aim to provide an adaptive control method for an aircraft and a computing device.

[0006] One aspect of this specification provides an adaptive control method for an aircraft, which is executed by an adaptive control system. The adaptive control system includes an outer loop controller, an inner loop controller, an adaptive controller, and a reference model. The reference model is a pre-trained machine learning model. The method includes:

[0007] Determine the real aircraft state and the simulated aircraft state corresponding to the current control cycle. The real aircraft state is determined by an on-board sensor, and the simulated aircraft state is determined by the reference model;

[0008] Using the adaptive controller, determine the control target compensation amount corresponding to the current control period according to the difference between the real aircraft state and the simulated aircraft state;

[0009] According to the control target compensation amount and the initial control target corresponding to the current control period, determine the corrected control target corresponding to the current control period, and the initial control target is determined using the outer loop controller;

[0010] Using the inner loop controller, control the aircraft according to the corrected control target and the real aircraft state, and use the reference model to simulate the aircraft state of the next control period of the current control period according to the corrected control target and the simulated aircraft state.

[0011] In some implementation manners, the aircraft state specifically includes: body-axis system speed, angular velocity, attitude angle, and earth-axis system speed.

[0012] In some implementation manners, the real aircraft state specifically includes: real aircraft position and real aircraft speed;

[0013] Before determining the corrected control target corresponding to the current control period, it further includes:

[0014] Using the outer loop controller, determine the initial target speed and initial target position corresponding to the current control period according to the real aircraft state and the preset flight target, and use them as the initial control target corresponding to the current control period.

[0015] In some implementation manners, using the inner loop controller to control the aircraft according to the corrected control target and the real aircraft state specifically includes:

[0016] Using the inner loop controller, determine the first throttle command corresponding to the current control period according to the corrected control target and the real aircraft state;

[0017] Control the aircraft according to the first throttle command.

[0018] In some implementation manners, it further includes:

[0019] Obtain a sample data set, where the sample data set includes the aircraft state and throttle command corresponding to each control period of the target aircraft;

[0020] For any control period, use the aircraft state and throttle command corresponding to this control period as the target sample;

[0021] Input the target sample into the reference model to be trained, and determine the predicted aircraft state change rate corresponding to the current control cycle output by the reference model to be trained;

[0022] Train the reference model to be trained according to the difference between the predicted aircraft state change rate and the true aircraft state change rate corresponding to the target sample.

[0023] In some implementation manners, use the reference model to simulate the aircraft state of the next control cycle of the current control cycle according to the corrected control target and the simulated aircraft state, specifically including:

[0024] Use the inner loop controller to determine the second throttle command corresponding to the current control cycle according to the corrected control target and the simulated aircraft state;

[0025] Use the reference model to determine the aircraft state change rate corresponding to the current control cycle according to the second throttle command and the simulated aircraft state;

[0026] Simulate the aircraft state of the next control cycle of the current control cycle according to the aircraft state change rate corresponding to the current control cycle and the simulated aircraft state.

[0027] In some implementation manners, the throttle command specifically includes: rotor thrust throttle command, pitch throttle command, and roll throttle command.

[0028] In some implementation manners, before obtaining the sample data set, it further includes:

[0029] Establish a simulation dynamics model corresponding to the target aircraft according to the control instruction module, motor speed controller model, aerodynamic model, dynamic equation, and kinematic equation corresponding to the target aircraft;

[0030] Use the simulation dynamics model to simulate the simulated aircraft state of the target aircraft under various preset working conditions;

[0031] Determine the sample data set corresponding to the target aircraft according to the simulated aircraft state and the working condition corresponding to the simulated aircraft state.

[0032] In some implementation manners, the aircraft state further includes: flight environment parameters.

[0033] The second aspect of this specification provides a computing device, including a memory and a processor. An executable code is stored in the memory. When the processor executes the executable code, the method described in the first aspect is implemented.

[0034] In the adaptive control solution of the aircraft provided in the embodiments of this specification, by using a pre-trained reference model in combination with model reference adaptive control, accurate control commands can be provided under large flight envelope and strongly coupled operating conditions, improving the anti-interference ability of the aircraft. Description of the Drawings

[0035] To more clearly illustrate the technical solutions in the embodiments of this specification, the following briefly introduces the drawings required for the description of the embodiments. Obviously, the drawings described below are only some embodiments recorded in this specification. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0036] Figure 1 It is a schematic structural diagram of an adaptive control system provided by an embodiment of the present invention;

[0037] Figure 2 It is a schematic flowchart of an adaptive control method for an aircraft provided by an embodiment of the present invention;

[0038] Figure 3 It is a schematic flowchart of a method for training a reference model provided by an embodiment of the present invention. Detailed Embodiments

[0039] To enable those skilled in the art to better understand the technical solutions in this specification, the following clearly and completely describes the technical solutions in the embodiments of this specification in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments in this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this specification.

[0040] Generally, in a flight control system, the controller can output a throttle command based on the state of the aircraft in the current control cycle and a preset flight target (the flight target here can be the target position, the expected speed to reach the target position, etc.), and the actuator can adjust the output power of each motor according to the throttle command, thereby achieving the adjustment of the attitude and speed of the aircraft. Specifically, the complete process of controlling the aircraft can be decomposed into several consecutive control cycles. In each control cycle, the on-board sensors determine the state of the aircraft, the control system processes the state of the aircraft to output a throttle command, and each actuator adjusts the output power of each motor according to the throttle command to complete the control of the aircraft in the current control cycle. Further, in the next control cycle of the current control cycle, the aircraft is continuously controlled according to the updated state of the aircraft. Among them, a control cycle can refer to the time required for the control system to complete a complete cycle from on-board sensor data acquisition, data processing to the actuator executing the throttle command. Generally, a control cycle of the flight control system is only 10 - 100 milliseconds.

[0041] However, the interference of environmental factors and the limited computing power of the controller make it difficult for a fixed-parameter controller to provide accurate throttle commands in a complex environment. In order to enable the aircraft to make more accurate adjustments according to the current state of the aircraft, the model reference adaptive flight control system in the prior art introduces a reference model and an adaptive controller. Among them, the reference model can determine the ideal aircraft state corresponding to any control cycle under a preset flight target, and the adaptive controller can, in the next control cycle of this control cycle, adjust the parameters of the controller according to the difference between the actual aircraft state after the execution of the throttle command in this control cycle and the ideal aircraft state corresponding to this control cycle, so that the parameters of the controller are adjusted to adapt to the aircraft state in each control cycle. Thus, the model reference adaptive flight control system has a stronger adaptability to different working conditions.

[0042] However, the reference model is also limited by the computing power of the on-board computing device of the aircraft. Generally, the reference model is generally designed based on a linear model or a low-order transfer function. The model reference adaptive flight control system built on this basis is still difficult to provide accurate throttle commands under large envelope (the change range of the aircraft state is large) or strong coupling (there is a strong mutual influence between different aircraft states, different throttle commands, and between the throttle command and the aircraft state) working conditions.

[0043] Therefore, Figure 1The structural schematic diagram of an adaptive control system in an embodiment of this specification is shown. The adaptive control system includes an outer loop controller, an inner loop controller, an adaptive controller, and a reference model. The adaptive control system can be deployed on an aircraft, send throttle commands to the actuators in the aircraft through the inner loop controller, and receive the aircraft states transmitted by the on-board sensors in the aircraft. It should be noted that, without special explanation, the "outer loop controller", "inner loop controller", "adaptive controller", and "reference model" involved in the following text all refer to the respective modules in the adaptive control system as Figure 1 shown.

[0044] Among them, the outer loop controller can be used to determine an initial control target according to the aircraft state. Here, the control target can refer to the target value of the aircraft state in the current control cycle; the reference model can be used to determine the simulated aircraft state in the current control cycle according to the simulated aircraft state and the simulated throttle command in the previous control cycle of the current control cycle. Here, the reference model can be a pre-trained machine learning model; the adaptive controller can be used to determine the control target compensation amount corresponding to the current control cycle according to the difference between the real aircraft state and the simulated aircraft state in the current control cycle; the corrected control target can be determined according to the initial control target and the control target compensation amount; the inner loop controller can be used to determine the throttle command according to the corrected control target; after the throttle command is input to the actuator, the control of the aircraft in one control cycle can be completed. Specifically, the aircraft state can include an outer loop state and an inner loop state. The outer loop state can include the observed values of the overall angles such as the position of the aircraft and the earth-axis system speed, and the inner loop state can include the attitude angle of the aircraft and the attitude angular velocity, etc., which can represent the aircraft attitude in a finer granularity.

[0045] It should be noted that Figure 1 the inner loop controller, outer loop controller, adaptive controller, etc. in the adaptive control system shown can be established by using various common control algorithms (for example, it can be a Proportion Integration Differentiation (PID) algorithm, an Active Disturbance Rejection Control (ADRC) algorithm, a Linear Quadratic Regulator (LQR) algorithm, a sliding mode control algorithm, a fuzzy control algorithm, etc.). This specification does not make any restrictions here.

[0046] Figure 2The flowchart of an adaptive control method for an aircraft in an embodiment of this specification is shown. The method is executed using an adaptive control system, which includes an outer loop controller, an inner loop controller, an adaptive controller, and a reference model. The reference model is a pre-trained machine learning model. This adaptive control system can be installed in the computing device of the aircraft or the remote control terminal of the aircraft, and this specification does not limit this. The method includes:

[0047] S201: Determine the real aircraft state and the simulated aircraft state corresponding to the current control period. The real aircraft state is determined using on-board sensors, and the simulated aircraft state is determined using the reference model.

[0048] As described above, the complete process of controlling the aircraft can be decomposed into several consecutive control periods. In each control period, the specific steps executed by the adaptive control system are the same. Only the control period - that is, the "current control period" - is taken as an example to introduce an adaptive control method for an aircraft provided in this specification.

[0049] First, the on-board sensors carried by the aircraft can collect the real aircraft state of the aircraft at the beginning of the control period The aforementioned on-board controller can include, but is not limited to, an acceleration sensor, a pressure sensor, a gyroscope sensor, etc. This specification does not limit the number and specific types of on-board sensors. The on-board sensors can output in real time the observed values related to the motion state or environmental information of the aircraft as the real aircraft state corresponding to the aircraft.

[0050] On the other hand, the adaptive control system also includes a reference model. In the previous control period of the control period , the reference model can determine the simulated aircraft state of the aircraft after receiving the simulated throttle command in the control period according to the simulated aircraft state at the beginning of the control period and the simulated throttle command corresponding to this control period .

[0051] In some implementation manners, the reference model can be a neural network model, such as a Deep Neural Networks (DNN), a Residual Network (ResNet), etc.

[0052] In some implementations, when training the reference model, the aircraft states and throttle commands under various working conditions can be prepared in advance as sample inputs, and the aircraft state after the aircraft receives the throttle command under ideal conditions can be determined as the sample label. Thus, the control period output by the trained reference model The corresponding simulated aircraft state can represent the ideal aircraft state of the aircraft at the beginning of the control period Herein, the aforementioned ideal conditions may mean that there are no abnormalities in the software and hardware devices of the aircraft and there are no external environmental disturbance factors.

[0053] In some implementations, when training the reference model, the aircraft state used as the sample input may only include the inner-loop state. Further, in the method as Figure 2 shown, the reference model only needs to predict the simulated aircraft state for the inner-loop state. Thus, the computational overhead and training overhead of the reference model can be reduced. On the other hand, since the output of the outer-loop controller for the initial control target does not require the assistance of the reference model, the reference model not predicting the outer-loop state does not affect the control accuracy of the adaptive control system.

[0054] S203: Using the adaptive controller, determine the control target compensation amount corresponding to the current control period according to the difference between the real aircraft state and the simulated aircraft state.

[0055] As described above, the simulated aircraft state can represent the ideal aircraft state at the beginning of the control period When there is a difference between the real aircraft state and the simulated aircraft state, generally speaking, it means that the aircraft is disturbed by external random disturbance amounts during flight - such as ground turbulence or gusts, etc., resulting in the throttle command in the control period not achieving the ideal control effect, and further causing the aircraft not to reach the ideal aircraft state at the beginning of the control period Hereby, after determining the real aircraft state and the simulated aircraft state, the adaptive controller can be used to determine the control target compensation amount corresponding to the control period

[0056]

[0057] to compensate for the control target that the aircraft fails to achieve due to external random disturbances in the control period.

[0057] It should be noted that different from the methods used in the prior art, in the method as Figure 2 shown, the output result of the adaptive controller is the control target compensation amount, rather than the updated controller parameters. Thus, in the method as Figure 2In the method shown, there is no need to adjust the parameters of the inner loop controller or the outer loop controller. By using the control target compensation amount to correct the initial control target output by the outer loop controller, each control period can be corrected. Thus, there is no need to analyze equations with high parameter dimensions and complex coupling relationships between parameters, such as dynamic equations, aerodynamic formulas, and kinematic equations. The nonlinear dynamic characteristics of the aircraft can be fitted through the reference model, avoiding the problem of insufficient model fitting ability when using a linear model or a ground connection transfer function. On the other hand, the reasoning process of the reference model does not need to involve a large number of nonlinear operations in the nonlinear dynamic equation, and the computing device of the aircraft can also carry the reference model.

[0058] S205: Determine the corrected control target corresponding to the current control period according to the control target compensation amount and the initial control target corresponding to the current control period, where the initial control target is determined by the outer loop controller.

[0059] After determining the control target compensation amount, the adaptive control system can use the control target compensation amount to correct the initial control target output by the outer loop controller and determine the control period. The corresponding corrected control target.

[0060] Among them, when performing step S201 - determining the control period After the corresponding true aircraft state, the outer loop controller can determine the initial control target corresponding to the control period according to the preset flight target and the outer loop state in the true aircraft state. The corresponding initial control target.

[0061] Specifically, the "control target" in the foregoing control target compensation amount, initial control target, and corrected control target may include control targets for each inner loop state in the aircraft state - that is, the next control period of this control period The next control period The target values of each inner loop state.

[0062] In some implementation manners, since each inner loop state can be represented in the form of a scalar, the corresponding corrected control target can be determined by directly adding the initial control target and the control target compensation amount.

[0063] S207: Use the inner loop controller to control the aircraft according to the corrected control target and the true aircraft state, and use the reference model to simulate the aircraft state in the next control period of the current control period according to the corrected control target and the simulated aircraft state.

[0064] After determining the corrected control target, the corrected control target and the inner loop state in the true aircraft state can be input into the inner loop controller, so that the inner loop controller outputs the control period. The corresponding real throttle command, and then each actuator can adjust the power of each motor according to the real throttle command output by the inner loop controller to complete the control cycle Control of the aircraft.

[0065] On the other hand, the inner loop state in the modified control target and the simulated aircraft state can also be input into the inner loop controller so that the inner loop controller outputs a control cycle The corresponding simulated throttle command, so that the reference model can be based on the control cycle Corresponding simulated aircraft status and simulated throttle command to determine the control cycle The corresponding simulated aircraft status.

[0066] In some implementations, the reference model can be based on the control cycle The corresponding inner loop state in the simulated aircraft state and the simulated throttle command determine the control cycle The corresponding simulated aircraft status.

[0067] Thus, each control cycle is executed successively, and the following can be achieved: Figure 2 An adaptive control method for an aircraft is shown.

[0068] like Figure 2 An adaptive control method for an aircraft is shown, which utilizes a pre-trained reference model combined with model reference adaptive control to provide precise control instructions under large envelope and strong coupling conditions, thereby improving the aircraft's anti-interference capability.

[0069] In some implementations, the inner ring state may include the body axis velocity , angular velocity , attitude angle .

[0070] in, They represent the forward, lateral and vertical velocities in the body axis system respectively; They represent the roll angular velocity, pitch angular velocity and yaw angular velocity under the body axis system respectively; They represent the roll angle, pitch angle and yaw angle respectively.

[0071] Accordingly, the “control target” in the control target compensation amount, the initial control target and the modified control target may include target values for the aforementioned inner loop states.

[0072] In some implementations, the outer loop control state may include the aircraft position and the aircraft speed; correspondingly, the real aircraft state may include the real aircraft position and the real aircraft speed; Figure 2Before the step S205 shown, it further includes using the outer loop controller to determine the initial target speed and the initial target position corresponding to the current control period according to the real aircraft state and the preset flight target, as the initial control target corresponding to the current control period.

[0073] Specifically, the real aircraft state may include the real outer loop state and inner loop state collected by the on-board sensors; the preset flight target can be determined according to the flight mission executed by the aircraft. In some implementation manners, the flight mission may include the preset final position of the aircraft and the final speed when reaching the final position. Before executing the step S205, the outer loop controller can use the real outer loop state and the preset flight target to determine the initial target speed and the initial target position corresponding to the current control period.

[0074] In some implementation manners, in Figure 2 the step S207 shown, using the inner loop controller, according to the corrected control target and the real aircraft state, determine the first throttle command corresponding to the current control period, and control the aircraft according to the first throttle command.

[0075] Specifically, the corrected control target and the inner loop state in the real aircraft state can be input into the inner loop controller, so that the inner loop controller can determine the control period according to the difference between the inner loop state in the real aircraft state and the corrected control target corresponding real throttle command - that is, the first throttle command. Since the control target compensation amount is referred to in the process of determining the corrected control target, each actuator uses the first throttle command to control the output power of each motor, which can achieve compensating for the external random disturbance received by the aircraft in the control period in the control period of the aircraft.

[0076] Figure 3 The figure shows a schematic flowchart of a method for training a reference model in an embodiment of this specification, including:

[0077] Step S301: Obtain a sample data set, where the sample data set includes the aircraft state and throttle command corresponding to each control period of the target aircraft.

[0078] Among them, the sample data set may include at least one set of aircraft states and throttle commands corresponding to a number of consecutive control periods.

[0079] In some implementation manners, the sample data set can be collected by outputting each on-board sensor and the inner loop controller of the target aircraft under the condition of no external random disturbance. It should be noted that the aircraft state in the sample data set can be the same as Figure 2The types of the observed values included in the inner-loop state in the method shown are the same. Correspondingly, the throttle command in the sample dataset can be the same as Figure 2 the type of command included in the throttle command in the method shown.

[0080] In some implementations, the aircraft state in the sample dataset may only include body-axis system velocity , angular velocity and attitude angle , that is, the aircraft state in the sample dataset may not include yaw angle and yaw angular velocity. Correspondingly, in the method shown in Figure 2 , the simulated aircraft state output by the reference model may not include the simulated aircraft state for the yaw angle and yaw angular velocity. Since the influence of the yaw angle on the overall aircraft attitude is weaker than that of other dimensions, the computational cost and training cost of the reference model can be further reduced with less impact on the prediction accuracy.

[0081] Step S303: For any control period, use the aircraft state and throttle command corresponding to this control period as the target sample.

[0082] During a round of training, the aircraft state and throttle command corresponding to any control period - for example, the control period - can be used as the target sample.

[0083] In some implementations, according to the difference between the aircraft state of this control period and the aircraft state of the next control period of this control period , and the length of the control period, determine the true aircraft state change rate corresponding to this control period as the sample label corresponding to this control period .

[0084] Step S305: Input the target sample into the reference model to be trained, and determine the predicted aircraft state change rate corresponding to this control period output by the reference model to be trained.

[0085] After determining the target sample, input the target sample into the reference model to be trained. Use the output result of the reference model as the predicted aircraft state change rate corresponding to this control period.

[0086] Step S307: Train the reference model to be trained according to the difference between the predicted aircraft state change rate and the true aircraft state change rate corresponding to the target sample.

[0087] After determining the predicted aircraft state change rate, the reference model to be trained can be trained using a preset loss function according to the difference between the sample label and this predicted aircraft state change rate.

[0088] Specifically, the preset loss function can be a commonly used loss function such as the cross-entropy loss function, cosine similarity loss, etc., and this specification does not limit it here.

[0089] In some implementation manners, a test data set can also be preset. After training the reference model to be trained, the prediction accuracy of the reference model is tested using the test data set. When the prediction accuracy of the reference model is lower than the preset prediction accuracy, the reference model is continuously trained using the sample data set.

[0090] Among them, a set of input-output pairs in the test data set can be composed of the aircraft state corresponding to a control cycle and the true aircraft state change rate corresponding to this control cycle; when the difference between the predicted aircraft state change rate output by the reference model for the aircraft state corresponding to a control cycle and the true aircraft state change rate corresponding to this control cycle is less than the preset error threshold, it can be considered that the prediction result of the reference model for this input-output pair is accurate; according to the prediction results of the reference model for several input-output pairs, the prediction accuracy of the reference model can be determined.

[0091] In some implementation manners, in Figure 2 shown in step S207 as Figure 2 shown, using the inner loop controller, according to the corrected control target and the simulated aircraft state, determine the second throttle command corresponding to the current control cycle, use the reference model, according to the second throttle command and the simulated aircraft state, determine the aircraft state change rate corresponding to the current control cycle, and simulate the aircraft state of the next control cycle of the current control cycle according to the aircraft state change rate corresponding to the current control cycle and the simulated aircraft state.

[0092] Using the Figure 3 reference model trained by the method shown can output the corresponding aircraft state change rate according to the throttle command and the aircraft state. In some implementation manners, this reference model can be expressed by the following formula:

[0093]

[0094] Among them, can represent the parameters of this reference model. The left side of the equal sign is the output of this reference model, and the inside is the input of this reference model; has the same meaning as described above; represents the change rates corresponding to respectively; Indicates the throttle command, which are respectively the rotor thrust throttle command, the pitch throttle command, and the roll throttle command in sequence.

[0095] Thus, by inputting the second throttle command and the state of the simulated aircraft into the reference model, the control period output by the reference model can be obtained. The corresponding change rate of the simulated aircraft. Further, according to this control period The corresponding change rate of the simulated aircraft, the control period The initial state of the simulated aircraft and the length of the control period, the state of the simulated aircraft corresponding to this control period can be determined. The corresponding state of the simulated aircraft.

[0096] In some implementation manners, the throttle command specifically includes: the rotor thrust throttle command, the pitch throttle command, and the roll throttle command.

[0097] Correspondingly, by inputting each throttle command into the actuator for controlling the corresponding motor, the attitude and speed of the aircraft can be controlled according to the throttle command.

[0098] In some implementation manners, before step S301 as shown in Figure 3 , according to the control instruction module, motor speed controller model, aerodynamic model, dynamic equation, and kinematic equation corresponding to the target aircraft, a simulation dynamic model corresponding to the target aircraft is established. Using the simulation dynamic model, the simulation aircraft state of the target aircraft under various preset working conditions is simulated, and the sample data set corresponding to the target aircraft is determined according to the simulation aircraft state.

[0099] Since it is difficult to completely exclude the interference of external random disturbance amounts on the target aircraft in the real environment. The aircraft state of the target aircraft collected in the real flight mission is difficult to represent the ideal aircraft state of the target aircraft. Thus, a simulation dynamic model of the target aircraft can be established. Using this simulation dynamic model, the simulation aircraft state of the target aircraft near various preset working points is simulated. Further, the simulation aircraft state and throttle command corresponding to each working condition obtained by simulation are determined. The above-mentioned working conditions may include the aircraft state and flight target. For any working condition, the initial simulation aircraft state is determined according to this working condition, the throttle command corresponding to each control period is continuously determined according to the flight target, and the simulation aircraft state and throttle command corresponding to a series of consecutive control periods under this working condition can be determined according to the simulation dynamic model, the throttle command corresponding to each control period, and the initial simulation aircraft state.

[0100] Specifically, the simulation dynamic model can be established as follows:

[0101]

[0102] Among them, it can represent the aircraft state, including the body-axis system speed angular velocity , attitude angle and the earth-axis system speed ; It can include the rotor thrust throttle command , pitch throttle command , roll throttle command , yaw throttle command ; It can represent time; is the corresponding next moment - that is, the aircraft state in the next control period.

[0103] Among them, to establish a simulation dynamics model, various common simulation model design tools can be used, such as Simulink, FLIGHTLAB, JSBSim, etc., which are not limited in this specification.

[0104] In some implementation manners, the aircraft state can further include flight environment parameters.

[0105] Specifically, the flight environment parameters can include environmental parameters such as atmospheric pressure, temperature, humidity, etc. that may affect the aircraft control process.

[0106] Furthermore, in step S301 as shown in Figure 3 , the aircraft state in the obtained sample dataset can include flight environment parameters; in step S207 as shown in Figure 2 , the reference model can determine the simulation aircraft state corresponding to the control period according to the simulated throttle command, the inner-loop state in the simulated aircraft state, and the flight environment parameters .

[0107] It should be understood that descriptions such as "first" and "second" in this article are only used to distinguish similar concepts for the sake of simple description and do not have other restrictive effects.

[0108] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment.

[0109] The above description has been made of specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the acts or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0110] Those of ordinary skill in the art should further realize that the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of function in the above description. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those of ordinary skill in the art can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application. Among them, the software module can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the technical field.

[0111] The specific embodiments described above have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An adaptive control method for an aircraft, characterized in that The method is executed by using an adaptive control system, which includes an outer loop controller, an inner loop controller, an adaptive controller, and a reference model. The reference model is a pre-trained machine learning model. The method includes: Determine the real aircraft state and the simulated aircraft state corresponding to the current control cycle. The real aircraft state is determined by using on-board sensors, and the simulated aircraft state is determined by using the reference model; Use the adaptive controller to determine the control target compensation amount corresponding to the current control cycle according to the difference between the real aircraft state and the simulated aircraft state; According to the control target compensation amount and the initial control target corresponding to the current control cycle, determine the corrected control target corresponding to the current control cycle. The initial control target is determined by using the outer loop controller; Use the inner loop controller to control the aircraft according to the corrected control target and the real aircraft state, and use the reference model to simulate the aircraft state of the next control cycle of the current control cycle according to the corrected control target and the simulated aircraft state; Among them, using the inner loop controller to control the aircraft according to the corrected control target and the real aircraft state specifically includes: Use the inner loop controller to determine the first throttle command corresponding to the current control cycle according to the corrected control target and the real aircraft state; Control the aircraft according to the first throttle command; The method further includes: Obtain a sample data set, which includes the aircraft state and the throttle command corresponding to each control cycle of the target aircraft; For any control cycle, use the aircraft state and the throttle command corresponding to this control cycle as the target sample; Input the target sample into the reference model to be trained, and determine the predicted aircraft state change rate corresponding to the current control cycle output by the reference model to be trained; Train the reference model to be trained according to the difference between the predicted aircraft state change rate and the real aircraft state change rate corresponding to the target sample.

2. The method according to claim 1, characterized in that, The aircraft state specifically includes: body-axis system speed, angular velocity, attitude angle, and earth-axis system speed.

3. The method according to claim 1, wherein The real aircraft state specifically includes: real aircraft position and real aircraft speed; Before determining the corrected control target corresponding to the current control cycle, it further includes: Use the outer loop controller to determine the initial target speed and the initial target position corresponding to the current control cycle according to the real aircraft state and the preset flight target, and use them as the initial control target corresponding to the current control cycle.

4. The method according to claim 1, characterized in that Using the reference model to simulate the aircraft state of the next control cycle of the current control cycle according to the corrected control target and the simulated aircraft state specifically includes: Use the inner loop controller to determine the second throttle command corresponding to the current control cycle according to the corrected control target and the simulated aircraft state; Use the reference model to determine the aircraft state change rate corresponding to the current control cycle according to the second throttle command and the simulated aircraft state; Simulate the aircraft state of the next control cycle of the current control cycle according to the aircraft state change rate corresponding to the current control cycle and the simulated aircraft state.

5. The method according to claim 1, wherein The throttle command specifically includes: a rotor thrust throttle command, a pitch throttle command, and a roll throttle command.

6. The method according to claim 1, wherein Before obtaining the sample data set, it further includes: According to the control instruction module, motor speed controller model, aerodynamic model, dynamic equation, and kinematic equation corresponding to the target aircraft, establish the simulation dynamic model corresponding to the target aircraft; Use the simulation dynamic model to simulate the simulation aircraft state of the target aircraft under various preset working conditions; Determine the sample data set corresponding to the target aircraft according to the simulation aircraft state and the working condition corresponding to the simulation aircraft state.

7. The method according to claim 1, characterized in that The aircraft state further includes: flight environment parameters.

8. A computing device, comprising a memory and a processor, wherein executable code is stored in the memory, characterized in that, When the processor executes the executable code, the method described in any one of claims 1-7 is implemented.

Images