Attitude tracking control method for real-time updating compensation of unmanned aerial vehicle

Abstract

The invention discloses a tracking control method for real-time updating compensation of the attitude of an unmanned aerial vehicle. The tracking control method comprises an unmanned aerial vehicle Hammerstein model, an event trigger controller, an input compensation controller and an unmanned aerial vehicle attitude controller. The method comprises the following steps: firstly, establishing a Hammerstein model to carry out data processing on a nonlinear part and a linear part of an unmanned aerial vehicle, fitting a Hammerstein inverse model, and then combining an input compensation controller and an attitude PID controller to design an attitude composite controller; and designing and optimizing an event trigger controller according to an error between unmanned aerial vehicle sensor feedback and attitude expectation input to control the updating frequency of the unmanned aerial vehicle attitude composite controller. The method has the advantages that the unmeasurable part in the unmanned aerial vehicle rigid body dynamic model is compensated, the influence of the nonlinear part in the system on control can be reduced, the calculation consumption of the controller can be reduced by using the event triggering control strategy, the unmanned aerial vehicle regulation and control frequency is reduced, the attitude error range in the unmanned aerial vehicle flight process is controlled, and the control precision is improved. And the energy loss is reduced.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle control, and in particular to an unmanned aerial vehicle model compensation composite controller event-triggered attitude tracking control method. Background Art

[0002] With the continuous development of UAV control technology, designing UAV flight attitude control methods for various complex operations has become a research hotspot in the field of UAV control. Domestic and foreign scholars have conducted extensive research on the attitude control methods of quadrotor UAVs and proposed a variety of control strategies, such as PID control method, sliding mode control method, backstepping method, etc. With the continuous improvement of UAV flight mission requirements, UAV flight attitude control faces the following challenges: (1) system model uncertainty, with the characteristics of strong coupling, strong nonlinearity and strong time-varying, and (2) the existence of external interference in the system. Therefore, it is very necessary to study the UAV flight attitude tracking control method.

[0003] However, the above control strategies are all continuous algorithms, which require real-time calculation of the control algorithm, resulting in real-time adjustment in the actual UAV system, wasting computing resources and energy consumption. In addition, the actual UAV system control part is an embedded data controller, which needs to discretize the above continuous algorithm before running, resulting in the controller increasing the update frequency in order to maintain the stability of the tracking algorithm. On the other hand, high-frequency update output to the actuator will reduce the service life of the actuator. During the use of the UAV, the parameters of the actuator need to be recalibrated and modified, which increases the system workload and reduces efficiency. Therefore, an event-triggered tracking control method for real-time compensation of UAV attitude is needed. Summary of the invention

[0004] The invention comprises four parts, namely, a Hammerstein model of a UAV, an event triggering controller, an input compensation controller and a UAV attitude controller. Firstly, a Hammerstein model of a nonlinear power output-linear sensor solution of a quad-rotor UAV is established to collect and analyze data of a nonlinear part and a preceding part of the UAV, and a Hammerstein inverse model is fitted accordingly, thereby designing an input compensation controller. Then, a UAV attitude composite controller is designed by combining the input compensation controller with a UAV attitude PID controller, and an event triggering controller is designed and optimized according to the relationship between UAV sensor feedback data and the expected input of the UAV attitude. Then, an optimized event triggering control strategy is adopted to control the update frequency of the UAV attitude composite controller, so as to reduce the computing power consumption of the UAV flight control while meeting the flight requirements of the UAV.

[0005] In the present invention, the Hammerstein model of the nonlinear power output-linear sensor solution of the quad-rotor UAV is established to collect and analyze data of the nonlinear part and the linear part of the UAV;

[0006] The quad-rotor UAV system can be described by a nonlinear link and a linear link. The link from the input control of the quad-rotor UAV to the attitude change has the characteristics of strong coupling and instability, and is a nonlinear link. The attitude solution link can be regarded as a dynamic linear link. According to the above characteristics of the UAV, a nonlinear link and a linear link are used to form the UAV Hammerstein system model, and the model is as follows:

[0007] The nonlinear model expression of the UAV Hammerstein system model is established: , in, U ( G ( t ), A ( t ), F ( t )) for drones t The input of the nonlinear link at time B (·) is the X-type quadrotor drone direction matrix, M are the four lift inputs of the quadrotor system, l is the weight vector.

[0008] Establish the linear model expression of the UAV Hammerstein system model: , in, Q is the calibration coefficient, Q ∈ R , offset ( t 0 ) is the sensor initial calibration matrix, C for X Vector normalization coefficients, X ( t )for t The attitude matrix of each axis at each moment.

[0009] The UAV Hammerstein system model is described as: , in, D is an unmeasurable transformation matrix, d ( t ) is noise.

[0010] In the present invention, the input compensation controller includes two modules: Hammerstein linear part inverse model and Hammerstein nonlinear part inverse model calculation. The linear part inverse model calculation is based on the attitude solution link principle, and the linear part matrix of the Hammerstein system model is inverted to calculate the inverse model of the linear part model of the system; the nonlinear part inverse model calculation is to use the input of the linear model, combined with the system control input, and the LM algorithm is used to fit the nonlinear model, and the fitted model is inversely operated to calculate the inverse model of the nonlinear part;

[0011] Using the Hammerstein model of the quadcopter drone, the Hammerstein inverse model is solved by step-by-step fitting. First, the linear part of the system model is inverted, and then the nonlinear part of the system model is inverted. The two inverse models are arranged into L -1 ()- H -1 ().

[0012] L -1 () part needs to be modified in the original Hammerstein model L () part can be directly obtained by inverse calculation, then L -1 () Partial expression is:

[0013] Hammerstein system H () Some of them have unmeasurable matrices D , the LM algorithm is used to fit the nonlinear model, and the linear part is inverted to obtain , solve the output of the nonlinear model, namely:

[0014] according to H () Some of the characteristics can H () The model is expressed as:

[0015] Transforming the above formula, we can get:

[0016] At this time, you can H () is transformed into the case without considering the initial condition and The nonlinear relationship between H ()Confirm input With the estimated output , at this timeD Describe:

[0017] There are ( n a + n b +1) unknown number, at this time H The expression of () is described as:

[0018] right Perform LM optimization and define the objective function f , solve the following optimization problem: , in, is the objective function, there exists ( n a + n b +1) unknowns, Ω is the performance index. To minimize Ω, it is necessary to calculate the Hessian matrix H and matrix g of the objective function, iterate, and finally when the objective function f Satisfy the iteration termination conditions and obtain the best fitting model ;final H -1 ()The model fitting expression is:

[0019] The update frequency of the drone Hammerstein inverse model is directly related to the drone sensor solution frequency. The inverse operation of the sensor data collected in the drone Hammerstein model is obtained. , through the model control input u out ( t ) and the nonlinear part to get the output Fit the nonlinear model and then solve the inverse model of the nonlinear part. The fitted nonlinear model part includes factors such as environment and interference. The matrix parameters need to be updated in real time to maintain the robustness of the UAV flight control system.

[0020] In the present invention, the UAV attitude composite controller includes two modules: a UAV attitude PID controller and an input compensation controller. The UAV attitude PID controller includes an attitude angle control outer loop and an attitude angular velocity control inner loop. The input of the input compensation controller is the expected attitude angle of the UAV. The inner loop and outer loop compensation control quantities of the UAV attitude PID controller are calculated respectively through the Hammerstein inverse model to reduce the interference of the nonlinear part entering the control system.

[0021] definition f , i , ψ Respectively represent the attitude angles in the inertial coordinate system: f is the roll angle, i is the pitch angle, ψ is the yaw angle, and the desired attitude angle vector is F d =[ f d i d ψ d ] T , the attitude angle sensor feedback is obtained as F t =[ f t i t ψ t ] T , the attitude angular velocity sensor feedback obtained is ; The attitude angle error is F e = F t- F d , define the attitude angular velocity error as W e = W t - u Фout ( t ), define the UAV attitude controller as a dual-loop PID controller, combine the input compensation controller with the PID controller, design the UAV attitude composite controller, and the quadrotor UAV attitude subsystem controller expression: , in, F e ( t )for t The error value between the attitude angle feedback data and the expected angle at the moment, W e ( t )for t The error value between the attitude angular velocity feedback data and the system angular velocity expectation at any moment;

[0022] UAV attitude composite controller attitude angle loop controller: , in, v d1 ( t) is the intermediate value of the compensation controller, is the estimated value of the compensation controller output, c 11 , c 12 is a constant coefficient;

[0023] UAV attitude composite controller attitude angular velocity loop controller: , in, v d2 ( t ) is the intermediate value of the compensation controller, is the estimated value of the compensation controller output, c 21 , c 22 is a constant coefficient.

[0024] In the present invention, the event-triggered controller designs a quadrotor UAV attitude expected input event-triggered control strategy based on the relationship between the UAV sensor feedback data and the UAV attitude expected input, adjusts the update frequency of the UAV attitude controller and meets the UAV flight requirements at the same time.

[0025] In order to solve the problems of large calculation amount and serious physical life loss caused by too frequent attitude control of drones, an event-triggered control strategy is introduced to control the update frequency of the drone attitude PID controller, so as to ensure the attitude tracking effect when the attitude algorithm of drones is frequently updated. The specific solutions are as follows:

[0026] For event-triggered control condition detection, determine whether the motion state error meets the control state update condition; set the event-triggered control parameters, let { t Ф,k} represents the system trigger time sequence; when the system meets the control state update condition, the controller input of the subsystem is updated. The event trigger controller is described as: , e 1 ( t )= F e ( t ) e 2 ( t )= W e ( t )+ this 1 ( t ) wФ ( t )= U ( t )- or Ф tanh( or Ф e 2 ( t ) / e Ф ) e ( t )= w Ф ( t )- U ( t ) in, e 1 ( t )for t Atmosphere angle error information at all times, e 2 ( t )for t The error information of attitude angle and attitude angular velocity is included at all times. c is a constant, w Ф ( t ) is an event trigger detection function designed based on the controller of the present invention, or Ф , e Ф is the parameter vector;

[0027] Based on the fixed threshold event trigger strategy, the trigger condition expression is: t Ф,k+1 =inf{ t ∈ R || w Ф ( t )- U ( t )|≥ or Ф} in, t Ф,k+1 It is the controller update time, and constitutes the controller execution time interval set { t Ф,k+1 - t Ф,k}, k ∈ N + The lower bound of t* ;

[0028] Compared with the prior art, the benefits of the present invention are: compensating for the unmeasurable parts in the rigid body dynamics model of the drone, and utilizing event-triggered control strategies to reduce the computational consumption of the controller, reduce the frequency of drone control, control the range of attitude errors during drone flight, and reduce energy loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a block diagram of the control method of the present invention.

[0030] Figure 2 This is a block diagram of the UAV attitude composite controller of the present invention. DETAILED DESCRIPTION

[0031] The technical solution of the present invention is clearly and completely described below with reference to the accompanying drawings and embodiments.

[0032] See attached figure, Figure 1 This is a block diagram of the control method of the present invention. Figure 2 The block diagram of the UAV attitude composite controller of the present invention includes four parts: a UAV Hammerstein model, an event trigger controller, an input compensation controller, and a UAV attitude controller. First, a Hammerstein model of a quad-rotor UAV nonlinear power output-linear sensor solution is established to collect and analyze data of the nonlinear part and the linear part of the UAV, and a Hammerstein inverse model is fitted based on the Hammerstein model to design an input compensation controller. Then, the UAV attitude composite controller is designed by combining the input compensation controller with the UAV attitude PID controller, and an event trigger controller is designed and optimized according to the relationship between the UAV sensor feedback data and the UAV attitude expected input. Then, an optimized event trigger control strategy is used to control the update frequency of the UAV attitude composite controller, so as to reduce the UAV flight control computing power consumption while meeting the UAV flight requirements.

[0033] This embodiment adopts an X-type quad-rotor drone structure. According to the structural characteristics of the drone system, it can be described by a nonlinear link and a linear state. The link from the input control of the quad-rotor drone to the attitude change has the characteristics of strong coupling and instability, and is a nonlinear link. The attitude solution link can be regarded as a dynamic linear link. See attached figure, Figure 1 This is a block diagram of the control method of the present invention, in which the UAV rigid body power output is a nonlinear part, which is expressed as H (), the UAV sensor solution is the linear part, expressed as L (); Collect and analyze data on the nonlinear and linear parts of the UAV quadrotor to prepare for model fitting.

[0034] The nonlinear model expression of the UAV Hammerstein system model is established: , in, U ( G ( t ), A ( t ), F ( t )) for drones t The input of the nonlinear link at time B (·) is the X-type quadrotor drone direction matrix, M are the four lift inputs of the quadrotor system, l is the weight vector.

[0035] Establish the linear model expression of the UAV Hammerstein system model:

[0036] in, Q is the calibration coefficient matrix, Q ∈ R , offset ( t 0 ) is the sensor initial calibration matrix, C for X Vector normalization coefficients, X ( t )for t The attitude matrix of each axis at all times.

[0037] The drone attitude sensor includes a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer. X The state includes three-axis acceleration, three-axis angular velocity and three-axis angle. offset ( t 0 ) is the initial calibration vector of the accelerometer, gyroscope and magnetometer;

[0038] Q The calibration coefficient expression is:

[0039] in, E 1 , E 2 , E 3 is a 3*3 identity matrix, m 1 , m 2 ,m 3 is the linear transformation coefficient;

[0040] C The normalized expression of is:

[0041] The UAV Hammerstein system model is described as: , in, D is an unmeasurable transformation matrix, d ( t ) is noise.

[0042] See attached figure, Figure 2 This is a block diagram of the UAV attitude composite controller of the present invention, where L -1 The () part is the inverse model of the linear part. H -1 The () part is a nonlinear partial inverse model;

[0043] It is easy to obtain that the linear part matrices are all reversible matrices. L -1 () part can be obtained by directly inverting the original linear model, then L -1 () Partial expression is:

[0044] Hammerstein system H () uses the LM algorithm to fit the nonlinear model, and obtains the , solve the output of the nonlinear model, namely:

[0045] according to H () Some of the characteristics can H () The model is expressed as:

[0046] Transforming the above formula, we can get:

[0047] At this time, you can H () is transformed into the case without considering the initial condition and The nonlinear relationship between H ()Confirm input With the estimated output , at this timeD Describe:

[0048] There are ( n a + n b +1) unknown number, at this time H The expression of () is described as:

[0049] right Perform LM optimization and define the objective function f , solve the following optimization problem: , in, is the objective function, there exists ( n a + n b +1) unknowns, Ω is the performance index. To minimize Ω, it is necessary to calculate the Hessian matrix H and matrix g of the objective function, iterate, and finally when the objective function f Satisfy the iteration termination conditions and obtain the best fitting model ;final H -1 ()The model fitting expression is:

[0050] In this embodiment, the input compensation controller is composed of a Hammerstein inverse model as a main component, and the parameters of the Hammerstein inverse model are updated according to the input and output of the Hammerstein model;

[0051] The update frequency of the drone Hammerstein inverse model is directly related to the drone sensor solution frequency. The inverse operation of the sensor data collected in the drone Hammerstein model is obtained. , through the model control input u out ( t ) and the nonlinear part to get the output Fit the nonlinear model and then solve the inverse model of the nonlinear part. The fitted nonlinear model part includes factors such as environment and interference. The matrix parameters need to be updated in real time to maintain the robustness of the UAV flight control system.

[0052] f , i , ψ Respectively represent the attitude angles in the inertial coordinate system: f is the roll angle, i is the pitch angle, ψ is the yaw angle; the desired attitude angle vector is defined as F d =[ f d i d ψ d ] T , the attitude angle sensor feedback is obtained as F t =[ f t i t ψ t ] T , the attitude angular velocity sensor feedback obtained is ; The attitude angle error is F e = F t- F d , the attitude angular velocity error is W e = W t - u Фout ( t ), the UAV attitude controller is defined as a dual-loop PID controller, the input compensation controller is combined with the PID controller, and the UAV attitude composite controller is designed. The expression of the quadrotor UAV attitude subsystem controller is: , in, F e ( t )for t The error value between the attitude angle feedback data and the expected angle at the moment, W e ( t )for t The error value between the attitude angular velocity feedback data and the system angular velocity expectation at any moment;

[0053] UAV attitude composite controller attitude angle loop controller: , in, v d1 ( t ) is the intermediate value of the compensation controller, is the estimated value of the compensation controller output, c 11 ,c 12 is a constant coefficient, c 11 , c 12 The values ​​of are respectively related to the correction of composite parameters of the outer loop of the UAV flight control attitude;

[0054] UAV attitude composite controller attitude angular velocity loop controller: , in, v d2 ( t ) is the intermediate value of the compensation controller, is the estimated value of the compensation controller output, c 21 , c 22 is a constant coefficient, c 21 , c 22 The values ​​are related to the composite correction parameters of the inner loop of the UAV flight control attitude.

[0055] In this embodiment, the event trigger controller detects the event trigger control condition and determines whether the motion state error meets the control state update condition; sets the event trigger control parameter, so that { t Ф,k} represents the system trigger time sequence; when the system meets the control state update condition, the controller input of the subsystem is updated. The event trigger controller is described as: , e 1 ( t )= F e ( t ) e 2 ( t )= W e ( t )+ this 1 ( t ) w Ф ( t )= U ( t )- or Ф tanh( or Ф e 2 (t ) / e Ф ) e ( t )= w Ф ( t )- U ( t ) in, e 1 ( t )for t Atmosphere angle error information at all times, e 2 ( t )for t The error information of attitude angle and attitude angular velocity is included at all times. c is a constant, w Ф ( t ) is an event trigger detection function designed based on the controller of the present invention, or Ф , e Ф is the parameter vector;

[0056] Based on the fixed threshold event trigger strategy, the trigger condition expression is: t Ф,k+1 =inf{ t ∈ R || w Ф ( t )- U ( t )|≥ or Ф} in, t Ф,k+1 It is the controller update time, and constitutes the controller execution time interval set { t Ф,k+1 - t Ф,k}, k ∈ N + The lower bound of t * .

[0057] When designing event-triggered control strategies, we should pay attention to avoiding the Zeno effect, that is, the lower bound t * The Zeno effect can be avoided. Now the control strategy of the present invention is verified:

[0058] For errore ( t ), and the derivative is:

[0059] The event-triggered control law can be obtained:

[0060] Obviously, in the above formula is a bounded function, that is, there must be a constant C Ф , so that ;exist t ∈[ t Ф,k , t Ф,k+1 ) Internal pair At the same time, we can get:

[0061] Combining the inequalities gives:

[0062] Therefore, there is t * Equivalent to t Ф,k+1 - t Ф,k ,get t * >0, so the time interval set { t Ф,k+1 - t Ф,k There is a lower bound t * , so that the system control model involved in the present invention can avoid the occurrence of Zeno effect.

[0063] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A tracking control method for real-time update compensation of UAV attitude, comprising four parts: UAV Hammerstein model, event trigger controller, input compensation controller, and UAV attitude controller; its characteristics are: Firstly, by establishing the Hammerstein model of the nonlinear power output-linear sensor solution of the quadrotor UAV, data collection and analysis are carried out on the nonlinear and linear parts of the UAV, and the Hammerstein inverse model is fitted based on it, and then the input compensation controller is designed; then the UAV attitude composite controller is designed by combining the input compensation controller with the UAV attitude PID controller, and the event trigger controller is designed and optimized according to the relationship between the UAV sensor feedback data and the expected input of the UAV attitude, and then the optimized event trigger control strategy is used to control the update frequency of the UAV attitude composite controller, so as to reduce the computing power consumption of the UAV flight control while meeting the flight requirements of the UAV.

2. The event-triggered tracking control method for real-time update compensation of the attitude of a quad-rotor drone according to claim 1 is characterized in that: The input compensation controller includes two modules: Hammerstein linear part inverse model and Hammerstein nonlinear part inverse model calculation. The linear part inverse model is based on the attitude solution link principle, and the linear part matrix of the Hammerstein system model is inverted to calculate the inverse model of the linear part model of the system; the nonlinear part inverse model uses the input of the linear model, combined with the system control input, and adopts the LM algorithm to fit the nonlinear model, performs inverse operation on the fitted model, calculates the inverse model of the nonlinear part, and updates the inverse model parameters according to the input and output of the Hammerstein system model.

3. The event-triggered tracking control method for real-time update compensation of the attitude of a quad-rotor drone according to claim 1 is characterized in that: The UAV attitude composite controller comprises two modules, namely, a UAV attitude PID controller and an input compensation controller. The UAV attitude PID controller comprises an attitude angle control outer loop and an attitude angular velocity control inner loop. The input of the input compensation controller is the expected attitude angle of the UAV. The inner loop and outer loop compensation control quantities of the UAV attitude PID controller are respectively calculated through the Hammerstein inverse model to reduce the interference of the nonlinear part entering the control system.

4. The event-triggered tracking control method for real-time update compensation of the attitude of a quad-rotor drone according to claim 1 is characterized in that: The event-triggered controller is designed based on the relationship between the feedback data from the drone sensor and the expected input of the drone attitude, and the quadcopter drone attitude expected input event trigger control strategy is adjusted to adjust the update frequency of the drone attitude controller while meeting the drone flight requirements.

Images