CN116301023A - Aircraft trajectory tracking method and device based on data-driven model predictive control - Google Patents

Abstract

The invention provides an aircraft trajectory tracking method and device based on data-driven model predictive control. The present invention first captures the latest aircraft state response data in real time during the flight of the aircraft, and uses the data-driven model predictive controller to optimize and solve the expected roll angle reference signal at the next moment as the input of the flight control of the aircraft to realize the plane of the aircraft. Horizontal trajectory tracking; at the same time, based on the real-time captured and stored aircraft state response data, online identification and dynamic adjustment of the model parameters used by the data-driven model predictive controller. The invention can capture the latest response data in real time during the flight of the aircraft, and identify and dynamically adjust the model parameters used by the model predictive controller on-line, so as to improve the flight robustness of the aircraft under different environmental working conditions. Based on the data-driven model predictive control method, the input and output data of the controlled system can be fully utilized to enhance the controller's ability to adapt to the system.

Description

Aircraft trajectory tracking method and device based on data-driven model predictive control

technical field

The invention relates to the field of aircraft trajectory tracking, in particular to an aircraft trajectory tracking method and device based on data-driven model predictive control.

Background technique

Thanks to the improvement of the computing power of terminal equipment, model predictive control has become one of the more common control strategies in the industry. The principle is to use the mathematical model to predict the future output according to the current system state and control input, and solve the optimal control sequence in the future, and repeat this process continuously for rolling optimization. Its characteristics include: it can deal with multivariable control problems, it can deal with input and output physical constraints, it can adapt to structural changes, etc. As a widely used control algorithm in the field of real-time control, model predictive control has close to optimal control effect in many cases.

With the increasing application of unmanned aerial vehicles with the advantages of low cost, small size, and easy maneuverability in the military and civilian fields, people have put forward higher requirements for the autonomous control capabilities of flight controllers. In the design of flight control systems at home and abroad The frequency of model predictive control technology is increasing year by year. However, in the existing relevant research results, people always only identify the system parameters once for the controlled system, and then take the obtained model as the inherent properties of the system and do not change any more. During the process, it may be necessary to face a changeable flight environment: under different operating conditions, the aerodynamic response parameters of the aircraft will change unpredictably, resulting in the aerodynamic characteristics of the aircraft in actual flight and the estimated value of aerodynamic calculation or ground wind. The difference in the measured values of the hole test degrades the performance of the control system designed based on the nominal dynamic model. Therefore, in order to achieve a stable and adaptive control algorithm, it is necessary to propose an idea similar to incremental training in the field of machine learning for traditional model predictive controllers, and provide a technical framework for dynamic model updates.

Contents of the invention

The object of the present invention is to solve the above-mentioned technical problems and provide a method and device for tracking aircraft trajectory based on data-driven model predictive control. The aerodynamic response characteristics of the aircraft are collected through pre-flight missions, the dynamic model is identified using the sparse identification method with control, and the model parameters of the model predictive controller are dynamically adjusted according to the real-time data of the flight process, so that the aircraft can complete better Target trajectory tracking task. The invention can make full use of the input and output data of the controlled system, and enhance the self-adaptive ability of the controller to the system.

The technical scheme that the present invention adopts is as follows:

An aircraft trajectory tracking method based on data-driven model predictive control, the method is: capturing and storing the latest aircraft state response data in real time during the flight of the aircraft; the aircraft state response data includes the aircraft's north position coordinates n, east Position coordinate e, yaw angle ψ _g , roll angle φ, roll rate p, lateral tracking error l _e and heading tracking error ψ _e ;

The aircraft state response data at the current moment is used as the input of the data-driven model predictive controller, and the data-driven model predictive controller is based on the aircraft's plane lateral dynamics equation, roll dimension response equation, lateral tracking error and heading tracking error equations to continuously predict the current situation in real time. Based on the aircraft state response data within a period of time after the current moment, the objective function is established based on the predicted aircraft state response data within a period of time after the current moment, and the optimal roll angle reference signal at the next moment is obtained. The expected roll angle reference signal is used to perform the flight mission, and realize the plane lateral trajectory tracking of the aircraft;

At the same time, the model parameters used by the data-driven model predictive controller are dynamically adjusted based on the online identification and dynamic adjustment of the aircraft state response data captured and stored in real time.

Further, the aircraft is a fixed-wing aircraft, and the plane transverse dynamics equation of the fixed-wing aircraft is expressed as:

n _k+1 =n _k +V _g cosψ _gk δt

e _k+1 ＝e _k +V _g sinψ _gk δt

ψ _gk+1 ＝ψ _gk +(gtanφ _k /V)δt

φ _k+1 ＝φ _k +p _k δt

The rolling dimension response equation is expressed as:

p _k+1 ＝p _k +(b ₀ φ _rk -a ₁ p _k -a ₀ φ _k )δt

The lateral tracking error equation is expressed as:

l _ek+1 ＝l _ek +V _g sinψ _g δt

The heading tracking error equation expresses:

ψ _ek+1 ＝ψ _ek +(gtanφ _k /V)δt

In the formula, n and e represent the northward position and eastward position of the aircraft respectively, both of which take the take-off coordinates as the origin; V _g represents the ground speed vector; ψ _g represents the plane angle between the ground speed vector and the north direction; g is the local gravitational acceleration; represents the roll angle; V represents the airspeed vector, p represents the roll angular velocity; φ _r is the expected roll angle reference signal input to the aircraft; a ₀ , a ₁ , b ₀ are the constant parameters of the model; k is the index of the sampling time, δt Indicates the time difference between sampling intervals.

Further, the model parameters used by the data-driven model predictive controller are pre-determined through the solution of the optimizer by using the collected flight data and using the sparse identification method with control.

Further, the flight data is collected by the following method: continuously input control excitation to the aircraft in a windless environment, the input control data includes multiple "2-1-1" double-cascade maneuvers with different amplitudes, and each input After the control reaches a steady state, the flight data is recorded, including the three-dimensional vector sequence of the roll angle, the roll rate, and the expected roll angle reference signal input to the aircraft.

Further, the data-driven model predicts that the expected roll angle reference signal output by the controller is set with hard constraints, so that the roll angle output by the controller is set within a safe range.

Further, the objective function includes the cost of lateral tracking error and heading tracking error, the penalty item of roll angle and the penalty item of roll maneuver.

Further, it also includes using a PID controller to track the vertical height of the aircraft.

An aircraft trajectory tracking device based on data-driven model predictive control, comprising:

The data acquisition and storage unit is used to capture and store the latest aircraft status response data in real time during the flight of the aircraft; the aircraft status response data includes the aircraft's northward position coordinate n, eastward position coordinate e, yaw angle ψ _g , Roll angle φ, roll rate p, lateral tracking error l _e and heading tracking error ψ _e ;

The data-driven model predictive controller takes the aircraft state response data at the current moment as input, and based on the plane lateral dynamics equation of the aircraft, the roll dimension response equation, the lateral tracking error and heading tracking error equations, it continuously predicts the time after the current moment in real time Based on the aircraft state response data within a period of time, the objective function is established based on the aircraft state response data within a period of time after the predicted current moment for optimization to obtain the optimal expected roll angle reference signal at the next moment, and the aircraft is based on the predicted expected roll angle The reference signal is used to perform flight missions and realize the plane lateral trajectory tracking of the aircraft;

The model parameter dynamic adjustment unit is used for online identification and dynamic adjustment of model parameters used by the data-driven model predictive controller based on the aircraft state response data stored in the data acquisition and storage unit.

Further, a PID controller is also included, which is used for longitudinal altitude tracking of the aircraft.

Further, in the PID controller for longitudinal altitude control, the target pitch angle output by the controller can be limited so that both the climb angle and the glide angle of the aircraft are less than 10 degrees.

The beneficial effects of the present invention are:

1. For the dynamic model of the aircraft, the present invention adopts the "2-1-1" maneuver combined with the sparse identification method to quickly identify the model parameters from the data, and has high accuracy and strong universality.

2. On the basis of the traditional model predictive controller, the present invention proposes an idea similar to incremental training in the field of machine learning, and provides a technical framework for dynamic update of the model, which can make full use of the input and output data of the controlled system to achieve more Robust adaptive control behavior.

3. The tracking control framework proposed by the present invention combines the data-driven model predictive controller with the PID controller, which can well complete the three-dimensional trajectory tracking task of the fixed-wing aircraft in three-dimensional space, and realize the spatial trajectory tracking of the aircraft The task can make the flight trajectory of the aircraft maintain a good convergence with the target path, and has excellent control performance.

4. The aircraft trajectory tracking method based on data-driven model predictive control of the present invention can be extended and applied to more control strategies.

Description of drawings

Fig. 1 is the control block diagram of the data-driven model predictive control method that the present invention is used for fixed-wing trajectory tracking task;

Fig. 2 is the schematic diagram that establishes aircraft dynamics model to be used coordinate system and parameter;

Fig. 3 is the relationship diagram between each software in the simulation environment;

Figure 4. Part of the dual-cascade "2-1-1" maneuver input and the corresponding output results of the aircraft roll in the flight data acquisition step;

Fig. 5 is a schematic diagram showing the relationship between the aircraft and the waypoint and the error term;

Fig. 6 is a comparison chart between the data restored by using the identified model and the actual flight data;

Fig. 7 is the comparison diagram of the data predicted using the identified model and the actual flight data;

Fig. 8 is the result figure of fixed-wing unmanned aerial vehicle frame tracking flight test;

Fig. 9 is the result diagram of the space trajectory tracking flight test of the fixed-wing UAV.

Detailed ways

The present invention provides an aircraft trajectory tracking method based on a data-driven model predictive control method, the method comprising: capturing and storing the latest aircraft state response data in real time during the flight process of the aircraft;

The aircraft state response data at the current moment is used as the input of the data-driven model predictive controller, and the data-driven model predictive controller is based on the aircraft's plane lateral dynamics equation, roll dimension response equation, lateral tracking error and heading tracking error equations to continuously predict the current situation in real time. Based on the aircraft state response data within a period of time after the current moment, the objective function is established based on the predicted aircraft state response data within a period of time after the current moment, and the optimal roll angle reference signal at the next moment is obtained. The expected roll angle reference signal is used to perform the flight mission, and realize the plane lateral trajectory tracking of the aircraft;

At the same time, based on the stored aircraft state response data, online identification and dynamic adjustment of the model parameters used by the data-driven model predictive controller are performed.

The tracking method proposed by the present invention constructs a data-driven model predictive controller based on the plane lateral dynamics equation of the aircraft, the roll dimension response equation, the lateral tracking error and heading tracking error equations, etc., and realizes the space trajectory tracking task of the aircraft, which can make The flight trajectory of the aircraft maintains good convergence with the target path, and has excellent control performance. The method of the present invention is applicable to various aircrafts. The fixed-wing aircraft is taken as an example below to describe the present invention in detail in conjunction with the accompanying drawings.

The right side of Fig. 1 is a kind of data-driven model predictive control method control block diagram for track tracking of fixed-wing aircraft provided by the present invention, and the realization of block diagram comprises the following steps:

以及机体坐标系/>

下的参量共同定义，平面横向动力学方程表示如下：Step S1, establishing the basic model of the model predictive controller. Firstly, the plane lateral dynamics equation of the aircraft is established. Figure 2 is a schematic diagram of the coordinate system and parameters used to establish the aircraft dynamics model. The plane lateral dynamics equations for fixed-wing aircraft can use a local inertial coordinate system consisting of north and east

and body coordinate system />

The following parameters are jointly defined, and the plane transverse dynamic equation is expressed as follows:

和/>

分别表示飞机的北向速度与东向速度，V_g表示地速矢量；ψ_g表示地速矢量与北向的平面夹角，以北向为基准顺时针旋转定义为正，取值范围[-π,π]；/>

表示地速矢量与北向的平面夹角的角速度，g为当地重力加速度；φ表示滚转角，以右滚定义为正；V表示空速矢量；/>

p表示滚转角速度。另外，图中的W表示风速。其次针对固定翼飞行器的滚转通道额外增加以下滚转维度响应方程：Among them, n and e represent the northward position and eastward position of the aircraft respectively, both of which take the take-off coordinates as the origin; the superscript · represents the derivative of the corresponding physical quantity,

and />

respectively represent the northward _speed and eastward speed of the aircraft, V _g represents the ground speed vector; ];/>

Indicates the angular velocity of the angle between the ground speed vector and the north plane, g is the local gravity acceleration;

p represents the roll angular velocity. In addition, W in the figure represents a wind speed. Secondly, for the roll channel of the fixed-wing aircraft, the following roll dimension response equation is additionally added:

是滚转角加速度；φ_r是输入飞行器的期望滚转角参考信号；a₀、a₁、b₀为模型的常数参量。in,

is the roll angle acceleration; φ _r is the expected roll angle reference signal input to the aircraft; a ₀ , a ₁ , b ₀ are the constant parameters of the model.

When the aircraft performs a trajectory tracking task, the target trajectory specified by the user is given in the form of a sequence of discrete waypoints. In the lateral trajectory tracking task, the waypoint information can be degenerated into two-dimensional plane coordinates, and it can be divided into KD tree space , and then extract the information of multiple waypoints starting from the nearest target point, project the plane track coordinate system of the aircraft, and perform high-order polynomial fitting under this axis system to obtain the curve F(x); Fig. 5 A schematic representation of the relationship between the aircraft and waypoints and the error term. Sub-figure (a) is after constructing a space-divided KD tree for the waypoint of the target trajectory, the program extracts the information of multiple waypoints starting from the nearest target point. Sub-figure (b) presents the curve F(x) obtained by high-order polynomial fitting of the selected waypoints in the plane track coordinate system of the UAV, and gives the lateral tracking error item l in the controller state quantity Geometric representation of _e and heading tracking error term _ψe .

Finally, based on the purpose of tracking, the specific expressions of the lateral tracking error and heading tracking error equations are established:

l _e ＝F(x)-y

Ψ _e = arctan(F′(x))-Ψ _g

The x and y in the formula are the spatial position parameters of the waypoint in the track coordinate system. F'(x) is the derivative of F(x).

Among them, the aircraft's north position coordinate n, east position coordinate e, collected yaw angle ψ _g , roll angle φ, roll rate p, lateral tracking error l _e and heading tracking error ψ _e are taken as state vectors, and the control The quantity is determined as the value of φ _r input to the flight control.

In step S3, in order to set the objective function in the computer and perform an optimal solution, it is necessary to discretize the previously given plane lateral dynamics equation, roll dimension response equation, lateral tracking error and heading tracking error equations of the aircraft into a differential form:

n _k+1 =n _k +V _g cosψ _gk δt

e _k+1 ＝e _k +V _g sinψ _gk δt

ψ _gk+1 ＝ψ _gk +(gtanφ _k /V)δt

φ _k+1 ＝φ _k +p _k δt

p _k+1 ＝p _k +(b ₀ φ _rk -a ₁ p _k -a ₀ φ _k )δt

l _ek+1 ＝l _ek +V _g sinψ _g δt

ψ _ek+1 ＝ψ _ek +(gtanφ _k /V)δt

Among them, k is the index of the sampling moment, and δt represents the time difference of the sampling interval.

Based on the above differential form, the data-driven model predictive controller can continuously predict the aircraft state response data within a period after the current moment in real time according to the input aircraft state response data at the current moment. Based on the state response data of the aircraft, the objective function is established for optimization, and the optimal expected roll angle reference signal at the next moment is obtained. The aircraft executes the flight mission according to the predicted expected roll angle reference signal, and realizes the plane lateral trajectory tracking of the aircraft.

Among them, the minimization of the objective function is a key part of the model predictive controller. This objective determines which control actions and flight states are desirable and which flight maneuvers are best avoided in order to better track the reference. In the horizontal trajectory tracking problem, the present invention provides a kind of optimization objective function:

ω_φr、/>

代表各项的权重系数，用于调节各项在最优化问题中的重要程度。N表示模型预测控制器的预测区间长度，式中第一行描述了横向跟踪误差与航向跟踪误差产生的代价，使其最小化能够让飞行器更加精准地跟踪参考轨迹；第二行描述了滚转角度的惩罚项，减小该项能够让飞行器在跟踪参考的同时尽量以平稳的姿态飞行；第三行描述了滚转机动惩罚项，最小化该项可以保证控制器输出控制量的过渡过程更加平滑，使飞行器在任务过程中避免出现频繁的大幅度机动，从而增加飞行稳定性、减小能量消耗。in

ω _φr 、/>

Represents the weight coefficient of each item, which is used to adjust the importance of each item in the optimization problem. N represents the length of the prediction interval of the model predictive controller. The first line of the formula describes the cost of the lateral tracking error and the heading tracking error. Minimizing it can make the aircraft track the reference trajectory more accurately; the second line describes the roll Angle penalty term, reducing this term can make the aircraft fly as smoothly as possible while tracking the reference; the third line describes the roll maneuver penalty term, minimizing this term can ensure that the transition process of the controller output control amount is more stable. Smooth, so that the aircraft avoids frequent large-scale maneuvers during the mission, thereby increasing flight stability and reducing energy consumption.

As a preferred solution, the weight of the objective function can take the following values:

After testing, the various indicators of the data-driven model predictive controller (MPC controller) can be adjusted under this weight coefficient. Minimizing the objective function can allow the aircraft to have better plane trajectory tracking capabilities with fewer and more peaceful maneuvers. The most The optimization solution step can use Ipopt open source solver for real-time solution.

Step S4, in the process of real-time aircraft trajectory tracking, according to the real-time captured and stored aircraft state response data, use the sparse identification method (Sparse Identification of Nonlinear Dynamics with Control, SINDYc) for identification and dynamic adjustment. Based on the foregoing description, the model parameters used by the data-driven model predictive controller are mainly the three constant parameters a ₀ , a ₁ , and b ₀ in the rolling dimension response equation. For the roll channel, the stored aircraft state response data contains a three-dimensional vector sequence of roll angle, roll rate and input roll angle reference value; in the configuration of the candidate library, step S1 has given a definite second-order model form- Roll the dimension response equation, so the corresponding basis function can be set according to this equation. Therefore, the specific parameter values corresponding to a ₀ , a ₁ , and b ₀ can be solved by the optimizer to realize online identification and dynamic adjustment.

Furthermore, before the initial use, the model parameters used by the data-driven model predictive controller can be pre-determined through the optimization solution by using the collected flight data and using the sparse identification method with control.

The flight data is collected by performing a program-controlled flight mission. The fixed-wing aircraft needs to continuously input control excitation and record all flight data in a windless environment. These data should include multiple 2-1-1 double-cascade maneuvers with different amplitudes. , the input control range should be controlled within an appropriate range. Fig. 4 is part of the dual-cascade "2-1-1" maneuver input and the corresponding output of the aircraft roll used in the flight data acquisition step. The "2-1-1" maneuver is an alternate pulse signal input with the same amplitude and a duty ratio of 2:1:1. The improved double-cascade "2-1-1" maneuver is used as a primary control in the present invention The input combination is indicated by the dotted line in the figure; the solid line in the figure is the response status of the fixed-wing UAV under the excitation input. It should be noted that this process can use actual flight acquisition or simulation acquisition. In this embodiment, the simulation platform is built using mainstream open source tool frames such as PX4, Gazebo, and ROS, which can realize multi-level monitoring of aircraft, flight control, airborne computer, and atmospheric environment. The real simulation of the object, in which the program control part is completed by the ROS node program, the block diagram of each software is shown in Figure 3. The platform realizes the real simulation of multiple objects such as aircraft, flight control, airborne computer and atmospheric environment. Among them, Gazebo is a set of robot simulation tools, with a built-in high-performance real-time physics engine, which provides real 3D physics calculation functions, and can create a virtual fixed-wing UAV in Gazebo, and configure it including wind field speed and turbulence in The dynamic behavior of the aircraft can be simulated through the simulator and the airborne sensor data with simulated noise can be output. PX4 is an open source unmanned system control software solution, which can access the sensors and actuators of the virtual aircraft and realize low-level control. As a high-level control loop, the data-driven model predictive controller proposed by the present invention will run on the Robot Operating System (ROS), which is a general robot development tool set, which can standardize and coordinate the controller and its various components. The operation of a subsidiary process, and use the MAVROS software package to connect it to the PX4 flight control software.

As an implementation, the sampling frequency in the flight data acquisition process can be set to 20Hz, wherein the control input includes "2-1-1" double-cascade maneuvers with different amplitudes from 0.1 to 0.8 radians and at intervals of 0.1 radians, After each control unit completes the input, a standby gap long enough to reach a steady state is arranged.

Based on the collected flight data, Sparse Identification of Nonlinear Dynamics with Control (SINDYc) is also used to identify the specific parameter values corresponding to a ₀ , a ₁ , and b ₀ through the optimizer, and then determine the Initial values for the model parameters used by the data-driven model predictive controller.

After determining the initial value, use the basic model to fit and restore the flight state of the aircraft under the same initial state and the same control sequence conditions in the data collection process, evaluate the error status with the real flight record, and pass the verification if the accuracy meets certain requirements, otherwise Check whether the flight data collection and identification process is improperly operated, and re-execute the relevant steps.

Fig. 6 is a comparison diagram between the data obtained by using the training set identification in the collected flight data to restore the training set and the actual flight data of the training set. The figure shows the fitting results of the model to the data set under the same control sequence and the same initial state input as the training set, where the solid line and dashed line are the original curve and the fitted curve, respectively. It is not difficult to find that the identification model can well fit and restore the roll angle φ item in the data set when the roll range is small, but when the roll range is larger, there is a certain decrease in fitting accuracy. It is caused by the inability of the low-order model to accurately restore high-frequency signals. Fortunately, the fitting results are fully applicable to the design of model predictive controllers, and UAVs pursue smooth flight. In actual flight, such high-frequency and large-scale maneuvers are often rare. In most cases, the control input will be Falling in a smaller range of amplitudes is more conducive to the prediction of the flight state.

Fig. 7 is a comparison chart of the data obtained by using the training set identification in the collected flight data to predict the test set and the actual flight data of the test set. In order to verify the generalizability of the fitted parameters, it is also advantageous to include in the test set some input forms other than the “2-1-1” maneuver, such as some additional task-guided procedures. The solid line in the figure shows the flight data during a random guidance mission, while the dotted line is the forward prediction result obtained after inputting the same control sequence as the flight process into the identification model. It can be seen from the observation that the model prediction results in the 100-second experimental test can fit the real flight data well, and there is no obvious prediction deviation even after the experimental period. In the process of model predictive control, it is usually only necessary to predict the state of the system within a few seconds after the current moment, and the performance of the model has far exceeded the demand index of the controller. At the same time, the experiment also proves the universality of model parameter identification based on the "2-1-1" double-cascade maneuver dataset.

In addition, the stored aircraft status response data can be continuously stored in the actual flight process by maintaining a queue whose length can be adjusted according to the actual situation, and it can be used as an incremental data set in the model identification process. By adjusting the The length of the queue can also control the correction weight of the incremental data set to the basic parameters of the model. Further, the incremental data set can be sparsely identified together with the previously obtained flight data collected by the "2-1-1" double-cascade maneuver when necessary, so that the model parameters can be corrected according to the latest data, and the model can be dynamically corrected. Realize data-driven dynamic model identification.

As an implementation, the length of the data queue for storing the real-time status and control input of the aircraft can be set to 10 seconds at a sampling rate of 20 Hz. By adjusting the length of the queue, the correction weight of the incremental data set to the basic parameters of the model can also be controlled.

Furthermore, hard constraints can be specified for the model predictive controller, which typically originate from the factory physical constraints of each component of the system and constraints based on safety considerations. The response of the aircraft lags behind the control quantity, and the corresponding state quantity can be constrained by restricting the control quantity. Here, it is not necessary to add any form of hard constraints to the state quantity of the aircraft, as long as it is for the stability of the aircraft flight and the safety of automatic driving Consider setting the roll constraint for the controller output.

As an implementation, the hard constraint on the output of the model predictive controller can set the roll angle between -40 degrees and 40 degrees.

Further, a PID controller can also be used to improve the longitudinal altitude control of the aircraft. Specifically, the multiple waypoint information obtained in the MPC controller starting from the nearest point of the current aircraft position is reused, and its height is interpolated and fitted, so as to calculate the smoothly changing target height reference value during the flight Enter the controller. Here, the output of the PID controller is set as the pitch angle reference value to approach the target height. For safety reasons, the target pitch angle output by the controller can be limited so that the climb angle and the glide angle of the aircraft are both less than 10 degrees; The segment function curve is used to obtain the appropriate throttle output setting value according to the pitch reference value mapping, and provide an energy management strategy for the system. Its throttle output value and pitch control value are sent to the flight controller to realize the longitudinal height tracking function.

Send a box trajectory tracking command to the aircraft. The side length of the box path is 100 meters, which is already a relatively small value for a fixed-wing aircraft and is quite challenging. Utilize the method of the present invention to carry out tracking, and use the power system sparse identification method with control (Sparse Identification of Nonlinear Dynamics with Control, SINDYc) to the model parameter used by the data-driven model predictive controller according to the aircraft state response data stored in the tracking flight test process Recognition, dynamic adjustment, the adjustment frequency is once every 10 seconds. Fig. 8 is a result diagram of the frame tracking flight test of the fixed-wing UAV. Sub-figure (a) is the control effect achieved by using the proposed control method of the present invention under the condition of no wind, sub-figure (b) is the tracking effect of the aircraft after the southwest average wind with a wind speed of 4 meters per second is added, and sub-figure (c ) is the trajectory tracking result under the southwest gust environment with an average wind speed of 4 m/s and a maximum wind speed of 8 m/s. It generally shows that the controller used has excellent performance and can well complete the plane trajectory tracking task, and can achieve more robust Great adaptive control behavior.

Fig. 9 is the result diagram of the space trajectory tracking flight test of the fixed-wing UAV. The space trajectory includes plane trajectory tracking commands and altitude tracking commands at the same time. The figure shows the differences between the plane coordinates, altitude coordinates, roll angle and the actual reference value during the mission execution process of the aircraft. The results show that the controller performs very well.

Corresponding to the aforementioned aircraft trajectory tracking method based on data-driven model predictive control, the present invention also provides an aircraft trajectory tracking device based on data-driven model predictive control.

An aircraft trajectory tracking device based on data-driven model predictive control provided by an embodiment of the present invention includes:

The data acquisition and storage unit is used to capture and store the latest aircraft status response data in real time during the flight of the aircraft; the aircraft status response data includes the aircraft's northward position coordinate n, eastward position coordinate e, yaw angle ψ _g , Roll angle φ, roll rate p, lateral tracking error l _e and heading tracking error ψ _e ;

The data-driven model predictive controller takes the aircraft state response data at the current moment as input, and based on the plane lateral dynamics equation of the aircraft, the roll dimension response equation, the lateral tracking error and heading tracking error equations, it continuously predicts the time after the current moment in real time Based on the aircraft state response data within a period of time, the objective function is established based on the aircraft state response data within a period of time after the predicted current moment for optimization to obtain the optimal expected roll angle reference signal at the next moment, and the aircraft is based on the predicted expected roll angle The reference signal is used to perform flight missions and realize the plane lateral trajectory tracking of the aircraft;

The model parameter dynamic adjustment unit is used for online identification and dynamic adjustment of model parameters used by the data-driven model predictive controller based on the aircraft state response data stored in the data acquisition and storage unit.

Further, a PID controller is also included, which is used for longitudinal altitude tracking of the aircraft.

The device embodiments can be realized by software, also by hardware or a combination of software and hardware.

As for the device embodiment, since it basically corresponds to the method embodiment, for related parts, please refer to the part description of the method embodiment. The device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network elements. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present invention. It can be understood and implemented by those skilled in the art without creative effort.

Apparently, the above-mentioned embodiments are only examples for clear description, rather than limiting the implementation. For those of ordinary skill in the art, other changes or changes in different forms can be made on the basis of the above description. It is not necessary and impossible to exhaustively list all implementation modes here. However, the obvious changes or variations derived therefrom still fall within the protection scope of the present invention.

Claims (9)

1. An aircraft trajectory tracking method based on data-driven model predictive control, characterized in that the method is: capturing the latest aircraft state response data in real time during aircraft flight; said aircraft state response data includes the northward position coordinates of the aircraft n, east position coordinate e, yaw angle ψ _g , roll angle φ, roll rate p, lateral tracking error l _e and heading tracking error ψ _e ; The aircraft state response data at the current moment is used as the input of the data-driven model predictive controller, and the data-driven model predictive controller is based on the aircraft's plane lateral dynamics equation, roll dimension response equation, lateral tracking error and heading tracking error equations to continuously predict the current situation in real time. Based on the aircraft state response data within a period of time after the current moment, the objective function is established based on the predicted aircraft state response data within a period of time after the current moment, and the optimal roll angle reference signal at the next moment is obtained. The expected roll angle reference signal is used to perform the flight mission, and realize the plane lateral trajectory tracking of the aircraft; At the same time, the model parameters used by the data-driven model predictive controller are dynamically adjusted based on the online identification and dynamic adjustment of the aircraft state response data captured and stored in real time. 2. method according to claim 1, is characterized in that, described aircraft is fixed-wing aircraft, and the plane transverse dynamics equation of fixed-wing aircraft is expressed as: n _k+1 =n _k +V _g cosψ _gk δt e _k+1 ＝e _k +V _g sinψ _gk δt ψ _gk+1 ＝ψ _gk +(gtanφ _k /V)δt φ _k+1 ＝φ _k +p _k δt The rolling dimension response equation is expressed as: p _k+1 ＝p _k +(b ₀ φ _rk -a ₁ p _k -a ₀ φ _k )δt The lateral tracking error equation is expressed as: l _ek+1 ＝l _ek +V _g sinψ _g δt The heading tracking error equation expresses: ψ _ek+1 ＝ψ _ek +(gtanφ _k /V)δt In the formula, n and e represent the northward position and eastward position of the aircraft respectively, both of which take the take-off coordinates as the origin; V _g represents the ground speed vector; ψ _g represents the plane angle between the ground speed vector and the north direction; g is the local gravitational acceleration; represents the roll angle; V represents the airspeed vector, p represents the roll angular velocity; φ _r is the expected roll angle reference signal input to the aircraft; a ₀ , a ₁ , b ₀ are the constant parameters of the model; k is the index of the sampling time, δt Indicates the time difference between sampling intervals. 3. The method according to claim 1, characterized in that, the model parameters used by the data-driven model predictive controller are pre-determined through the solution of the optimizer by using the collected flight data and using the sparse identification method with control. 4. The method according to claim 3, wherein the flight data is collected by the following method: continuously input control excitation to the aircraft in a windless environment, and the input control data includes a plurality of "2-1" of different amplitudes. -1" double cascade maneuver, recording flight data after each input control reaches a steady state, including a three-dimensional vector sequence of roll angle, roll rate, and desired roll angle reference signal input to the aircraft. 5. The method according to claim 1, wherein the data-driven model predicts that the expected roll angle reference signal output by the controller is provided with hard constraints, so that the roll angle output by the controller is set within a safe range . 6 . The method according to claim 1 , wherein the objective function includes costs generated by lateral tracking error and heading tracking error, a roll angle penalty term and a roll maneuver penalty term. 7 . 7. The method according to claim 1, further comprising using a PID controller to track the vertical height of the aircraft. 8. An aircraft trajectory tracking device based on data-driven model predictive control, characterized in that it comprises: The data acquisition and storage unit is used to capture and store the latest aircraft status response data in real time during the flight of the aircraft; the aircraft status response data includes the aircraft's northward position coordinate n, eastward position coordinate e, yaw angle ψ _g , Roll angle φ, roll rate p, lateral tracking error l _e and heading tracking error ψ _e ; The data-driven model predictive controller takes the aircraft state response data at the current moment as input, and based on the plane lateral dynamics equation of the aircraft, the roll dimension response equation, the lateral tracking error and heading tracking error equations, it continuously predicts the time after the current moment in real time Based on the aircraft state response data within a period of time, the objective function is established based on the aircraft state response data within a period of time after the predicted current moment for optimization to obtain the optimal expected roll angle reference signal at the next moment, and the aircraft is based on the predicted expected roll angle The reference signal is used to perform flight missions and realize the plane lateral trajectory tracking of the aircraft; The model parameter dynamic adjustment unit is used for online identification and dynamic adjustment of model parameters used by the data-driven model predictive controller based on the aircraft state response data stored in the data acquisition and storage unit. 9. The device according to claim 8, further comprising a PID controller for tracking the vertical height of the aircraft.

Images