CN115685760A - A four-rotor hybrid fault-tolerant control method and system for actuator faults - Google Patents

Abstract

The invention provides a four-rotor hybrid fault-tolerant control method and system for actuator faults, which comprises the steps of analyzing and modeling a four-rotor aircraft structural framework and a mathematical control model, and providing a flight control structure based on a double closed-loop control loop; carrying out fault-tolerant control by adopting position control based on a PID model algorithm and attitude control based on an adaptive hybrid controller; attitude control combines a learning-based data-driven approach with a PID control algorithm. In order to improve the robustness of a control strategy, the fault condition of an actuator is considered, an RL (return link) and a classical model PID (proportion integration differentiation) control algorithm are combined, uncertain parameters can be estimated in a self-adaptive manner by means of a robust learning algorithm in attitude control without knowing detailed fault information, and a more robust fault-tolerant controller is designed for a dynamic system under a fault. The position control and the attitude control reserve a robust method based on a model, can offset the adverse effect of unknown external interference, and ensure the stability and the safety of the control.

Description

A four-rotor hybrid fault-tolerant control method and system for actuator faults

technical field

The invention belongs to the control field of autonomous and safe driving of unmanned aerial vehicles, and specifically relates to a four-rotor hybrid fault-tolerant control method and system for actuator failure caused by motor or propeller loss.

Background technique

Quadrotors are playing an increasing role in various potential applications such as environmental monitoring and surveillance, search and rescue operations, coordination and formation control, and various military applications due to their simple structure, strong hovering capability, and low operating costs . However, many specific applications may make unmanned quadrotors enter cluttered and dangerous environments. At the same time, quadrotors are inherently unstable, highly nonlinear and other safety and stability issues. In this case, the reliability and safety of unmanned aerial vehicles Control is a very challenging task.

During the flight, the control parameters of the quadrotor may produce many uncertain factors with the change of the operating environment, and system failure is inevitable. Failures can be attributed to failure of sensors or actuators. Actuator failures mainly include motor failures or propeller failures. When the quadrotor performs high-load and long-distance missions, due to the low rotor redundancy, it is more susceptible to actuator failure, or fatigue fracture of the blades, which eventually causes the UAV to lose control or even crash. At the same time, the loss of actuator effectiveness will also deteriorate the system performance, which will cause structural parameter changes or internal uncertainties in the model, leading to catastrophic consequences. Therefore, it is crucial to develop a fault-tolerant control method for quadrotors in the event of actuator failure. This reliability and safety requirement is beyond what conventional control strategies can provide, driving active research on fault-tolerant control (FTC) of quadrotors.

Many research works have achieved autonomous flight of quadrotors in the event of rotor failure. According to the controller architecture, currently designed fault-tolerant controllers can be divided into two categories, one is the model-based controller, and the other is the data-driven learning method. Model-based controllers, designed according to the physical implications of fault dynamics, are a common and standard approach to controlling UAVs and are currently well developed. For example, using the traditional cascaded LQR controller can still maintain the stable control of the quadrotor in the case of losing one, two or even three motor rotors, proving the performance of the model-based design. But the performance and robustness of this approach largely depend on the accuracy and comprehensiveness of the system dynamic model, and this fault-tolerant control strategy requires the design of specific estimators and controllers to detect and recover from specific faults. If there are fault types other than those specified in the actual scene, the actuators often cannot operate as expected, resulting in insufficient fault tolerance of these traditional controllers.

To address this problem, learning-based fault-tolerant control methods have been developed that can learn a direct control mapping from UAV states to motor commands. The method allows the development of control strategies using the large amount of data obtained from the system, achieving state-of-the-art performance on many control tasks. However, this method usually relies on a large amount of data to capture the nonlinear dynamics of the system, and the cost of obtaining data is relatively high. Moreover, due to various reasons such as modeling errors and disturbances, it generally lacks stability guarantees and robustness analysis guarantees in the control process, because A neural network with uninterpretable features is fundamental to its structural composition.

Contents of the invention

Since the actuator is a key component to control the motion of the UAV, it can be directly connected to the control signal. The present invention provides a four-rotor hybrid fault-tolerant control method for actuator failures and other situations such as actuator failures and uncertain parameters during flight. and system.

To achieve the above object, the present invention adopts the following technical solutions:

A quadrotor hybrid fault-tolerant control method oriented to actuator faults, characterized in that it includes:

Based on the principle of quadrotor flight, establish the dynamic model of the quadrotor aircraft, establish the ground reference coordinate system and the body coordinate system to represent the quadrotor flight structure, and use the dynamic equation to represent the quadrotor flight motion and actuator failure model;

According to the quadrotor flight motion and actuator failure model, the flight control structure is designed based on the double closed-loop control loop, and the fault-tolerant control design is carried out for the flight control structure, and the position outer loop control based on the PID method and the adaptive attitude inner loop based on the RL algorithm are adopted. The loop control realizes the quadrotor hybrid fault-tolerant control.

In order to optimize the above technical solutions, the specific measures taken also include:

相关联。Further, the ground reference coordinate system and the body coordinate system are respectively {R}(O, x, y, z) and {R _b }(O _b , x _b , y _b , z _b ), where the ground reference coordinate The system belongs to the Cartesian coordinate system and is fixed on the earth. The coordinate origin 0 is any point on the ground, O _b is the center of mass of the quadrotor, {R _b } and {R} pass through the position vector p=[x, y, z] of the quadrotor ^T and attitude angle vector

Associated.

Further, the use of dynamic equations to represent the quadrotor flight motion is specifically as follows:

The translational motion equation of the quadrotor in the case of external disturbance is as follows:

分别表示沿x，y，z三轴的线加速度，m为四旋翼的质量，g为重力加速度，

θ，ψ分别表示滚转、俯仰、偏航的姿态角，u₁是四旋翼垂直z轴方向的控制力，

表示机体沿x，y，z三轴的线速度，k_x，k_y，k_z是x，y，z三轴的阻力系数，

表示外部干扰；In the formula,

Respectively represent the linear acceleration along the x, y, and z axes, m is the mass of the quadrotor, g is the acceleration of gravity,

θ, ψ represent the attitude angles of roll, pitch, and yaw respectively, and u ₁ is the control force of the quadrotor in the direction of the vertical z-axis,

Indicates the linear velocity of the body along the three axes of x, y and z, k _x , k _y , k _z are the resistance coefficients of the three axes of x, y and z,

Indicates external interference;

The rotational motion equation of the quadrotor is as follows:

分别表示滚转、俯仰、偏航的姿态角加速度，机体绕(x，y，z)三轴的角速度为

u₂，u₃，u₄分别为滚转、俯仰、偏航方向上的扭矩，k₂，k₃，k₄是三个力矩的阻力系数，I_x，I_y，I_z分别为机体绕x，y，z轴的转动惯量，I_r为单个电机的转动惯量，Ω_r＝Ω₁-Ω₂+Ω₃-Ω₄为整体的转子角速度，Ω₁，Ω₂，Ω₃，Ω₄分别表示四个转子的角速度，

表示外部干扰。In the formula,

Represent the attitude angular acceleration of roll, pitch, and yaw respectively, and the angular velocity of the body around the three axes (x, y, z) is

u ₂ , u ₃ , u ₄ are the torques in the direction of roll, pitch, and yaw respectively; k ₂ , k ₃ , k ₄ are the drag coefficients of the three moments; I _x , I _y , I _z are the body’s winding The moment of inertia of the x, y, and z axes, I _r is the moment of inertia of a single motor, Ω _r =Ω ₁ -Ω ₂ +Ω ₃ -Ω ₄ is the overall rotor angular velocity, Ω ₁ , Ω ₂ , Ω ₃ , Ω ₄ represent the angular velocities of the four rotors respectively,

Indicates external interference.

Further, the establishment process of the actuator fault model is as follows:

The dynamic equations for establishing the output of each controller of the quadrotor, the control input of the actuator and the rotor speed are as follows:

u ₁ =F ₁ +F ₂ +F ₃ +F ₄ (7)

u ₂ =(F ₄ -F ₂ )L (8)

u ₃ =(F ₃ -F ₁ )L (9)

u ₄ ＝(τ ₁ -τ ₂ +τ ₃ -τ ₄ ) (10)

和

表示第n个电机的产生的扭矩和推力，w_n为四个螺旋桨的转速，n＝1，2，3，4，b为推力因子，d为阻力因子；L为四旋翼中心到各螺旋桨中心的距离；In the formula, u ₁ is the control force of the quadrotor in the vertical z-axis direction, u ₂ , u ₃ and u ₄ are the torques in the roll, pitch and yaw directions respectively;

and

Indicates the torque and thrust generated by the nth motor, w _n is the speed of the four propellers, n=1, 2, 3, 4, b is the thrust factor, d is the resistance factor; L is the center of the four rotors to the center of each propeller distance;

The actuator fault model equation is as follows:

y(t)=u _f +τu(t) (11)

Among them, u(t) represents the control input [u ₁ , u ₂ , u ₃ , u ₄ ] at time t, τ, 0≤τ≤1 represents the actuator fault index, and u _f represents the additive fault value of four actuators The resulting vector:

u _f =[f ₁ , f ₂ , f ₃ , f ₄ ] (12)

In the formula, f ₁ , f ₂ , f ₃ , and f ₄ represent the unknown constant additive faults of the four actuators.

Further, the flight control structure includes an outer loop controller, an inner loop controller and sensors, wherein the outer loop is a position loop, which is used to control the position, acceleration and speed of the UAV, track the position error and minimize it and generate the required attitude angle; the inner ring is the attitude ring, which is used to adjust the attitude angle of the aircraft;

The outer loop controller generates the thrust to control the altitude according to the desired altitude, generates the expected pitch angle and the expected roll angle according to the difference between the expected position and the actual position and provides them to the inner loop controller; the inner loop controller inputs the expected pitch angle , the expected roll angle and the expected yaw angle, the expected yaw angle is calculated according to the expected position, the inner loop controller generates the motor control signal of the expected torque based on the error between the three expected angles and the measured angle; the sensor uses For the state measurement of the quadrotor aircraft, the measured value is provided as feedback to the outer loop controller and the inner loop controller.

Further, the position outer loop control based on the PID method adopts a PID controller based on a dynamic model, and the error is reduced to zero through the PID equation. The error is the difference between the expected value and the actual value. The PID equation is as follows:

In the formula, x(t) is the bottom control output of the quadrotor, t represents the time, K _p , K _I and K _D are the proportional, integral and differential gain control parameters respectively, and e(t) is the distance between the expected value and the actual value difference;

The RL algorithm-based adaptive attitude inner-loop control adopts an RL adaptive algorithm to adjust and update the control parameters and weights of the PID controller.

Further, the RL algorithm acts as an upper-layer controller to adjust the PID parameters (K _p , K _I , K _D ) and participation weights δ ⁱ of the lower-layer PID controller in real time according to the current state, and the lower-layer PID controller outputs x(t) To compensate the error e(t) generated during the control process.

其中，π表示策略函数，

分别表示第i个PID控制器对应的比例、积分和导数参数，

θ_d，

分别表示设置的期望俯仰角、期望滚转角和期望偏航角；critic网络提供用于训练更新actor信息。Further, the RL adaptive algorithm adopts deep deterministic policy gradient DDPG, and the action output a(t) of the actor network in the DDPG algorithm is the controller parameter and weight information:

Among them, π represents the policy function,

respectively represent the proportional, integral and derivative parameters corresponding to the i-th PID controller,

θ _d ,

Respectively represent the expected pitch angle, expected roll angle and expected yaw angle of the settings; the critic network provides actor information for training and updating.

The present invention also proposes a four-rotor hybrid fault-tolerant control system for actuator faults, which is characterized in that it includes:

The calculation unit is used to establish the dynamic model of the quadrotor aircraft based on the principle of quadrotor flight, establish the ground reference coordinate system and the body coordinate system to represent the quadrotor flight structure, and use dynamic equations to represent the quadrotor flight motion and actuator failure model;

The control unit is used to design the flight control structure based on the double closed-loop control loop according to the quadrotor flight motion and the actuator failure model, and to carry out the fault-tolerant control design for the flight control structure, using the position outer loop control based on the PID method and the RL algorithm The self-adaptive attitude inner loop control realizes the quadrotor hybrid fault-tolerant control.

The present invention also proposes a computer-readable storage medium storing a computer program, wherein the computer program enables the computer to execute the above-mentioned four-rotor hybrid fault-tolerant control method for actuator failure.

The beneficial effects of the present invention are:

(1) The present invention fully considers the failure situation of the actuator, and carries out mathematical modeling to the four-rotor flight control process and internal dynamic structure for the four-rotor flight principle; in order to realize the autonomous flight control of the four-rotor aircraft, set up its dynamic model ; In order to accurately represent the dynamic state of the quadrotor and the space movement of different modes, the inertial ground reference coordinate system and the body coordinate system are established to represent its flight structure. Finally, the simplified dynamic equations are used to express the quadrotor flight motion and actuator failure model;

(2) A flight control structure using a closed-loop control loop, this control method allows stabilizing the position and direction of the flight trajectory. The feedback information obtained by the sensor measurement of the closed loop reduces the model dependence on the control method, which helps to deal with unknown and uncertain disturbances in the environment;

(3) The flight control structure is designed as a double-loop nested loop, including a position control loop and an attitude control loop. The attitude control module is responsible for adjusting the flight attitude, the position control adjusts the position variable, and forms a nested feedback loop together, which can ensure that the control performance is good, the structure is clear, easy to implement, and it has good robustness and accuracy under disturbance;

(4) The PID algorithm with fixed parameters is used as the position control of the outer loop and the combination of RL algorithm and model PID control is used as the attitude control structure of the inner loop. This double closed-loop fault-tolerant control structure can effectively deal with actuator faults and ensure the closed-loop stability and robustness of the system. The RL learning algorithm helps to adaptively estimate uncertain parameters without knowing the detailed fault dynamics information, while the model control algorithm can be used to offset the adverse effects of unknown external disturbances, making it fault-tolerant and resistant to disturbances.

Description of drawings

Fig. 1 is a schematic diagram of quadrotor dynamics modeling according to an embodiment of the present invention.

Fig. 2 is a diagram of an additive fault model of an actuator according to an embodiment of the present invention.

FIG. 3 is a structural diagram of a flight control based on a double closed loop according to an embodiment of the present invention.

Fig. 4 is a scheme diagram of a hybrid adaptive fault-tolerant controller according to an embodiment of the present invention.

FIG. 5 is a structural diagram of a PID model algorithm according to an embodiment of the present invention.

FIG. 6 is a network structure diagram of a DDPG algorithm according to an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the scope of protection of this application.

In one embodiment, the present invention proposes a four-rotor hybrid fault-tolerant control method for actuator failure caused by motor or propeller loss, which mainly includes the following steps:

1. Modeling of Quadrotor Structural Framework and Flight Control

S1.1: Based on the principle of quadrotor flight, mathematically model the flight control process and internal power structure. In order to realize the autonomous flight control of the quadrotor aircraft, its dynamic model must be established first. Secondly, in order to accurately represent the dynamic state of the quadrotor and the space movement of different modes, the ground reference coordinate system and the body coordinate system are established to represent its flight structure.

The specific method of step S1.1 is as follows:

)，但只有四个独立的输入(总推力和三个力矩)。而四旋翼的输入控制是通过改变四个转子的速度，从而产生四个旋翼推力进行差分控制来实现的，每个转子会产生沿垂直体轴的定向的推力。如图1所示的四旋翼飞行器动力学模型，其中包含了两个用于研究系统运动的坐标系：地面参考坐标系{R}(O，x，y，z)，以及四旋翼固定机体框架坐标系{R_b}(O_b，x_b，y_b，z_b)，其中地面坐标系属于笛卡尔坐标系，固定于地球，坐标原点O可任选地面上某一点，O_b是四旋翼的质心。{R_b}与{R}通过位置向量p＝[x，y，z]^T和三个独立角度

向量相关联。这两个向量充分描述了四旋翼相对于地球框架的平移和旋转运动。在飞行过程中，改变四个旋翼的转速，四旋翼将产生各种飞行姿态，比如滚转、俯仰和偏航角姿态运动，使四旋翼向预定方向运动，完成飞行任务。图1中，前转子M1和后转子M3沿逆时针方向旋转，另两个转子M2和M4沿顺时针方向旋转。当转子M1和M3的速度保持不变时，通过改变转子M2和M4的速度来获得滚转

类似地，当转子M2和M4的速度保持相同时，通过改变转子M1和M3的速度来实现俯仰(θ)。偏航(ψ)通过增加(减小)转子M1和M3的速度以及降低(增大)转子M2和M4的速度来获得。A quadrotor is an underactuated system because it has six states (position p = [x, y, z] ^T : and attitude angle

), but only four independent inputs (total thrust and three moments). The input control of the quadrotor is realized by changing the speed of the four rotors, thereby generating the thrust of the four rotors for differential control, and each rotor will produce a thrust oriented along the vertical body axis. The quadrotor aircraft dynamics model shown in Figure 1 contains two coordinate systems used to study the motion of the system: the ground reference coordinate system {R} (O, x, y, z), and the quadrotor fixed body frame Coordinate system {R _b }(O _b , x _b , y _b , z _b ), where the ground coordinate system belongs to the Cartesian coordinate system and is fixed on the earth. The origin O of the coordinates can be selected from a certain point on the ground, and O _b is a quadrotor centroid. {R _b } and {R} pass the position vector p=[x, y, z] ^T and three independent angles

Vectors are associated. These two vectors fully describe the translational and rotational motion of the quadrotor relative to the earth frame. During the flight, by changing the rotation speed of the four rotors, the quadrotor will produce various flight attitudes, such as roll, pitch and yaw angle attitude movements, so that the quadrotor will move in a predetermined direction to complete the flight mission. In FIG. 1 , the front rotor M1 and the rear rotor M3 rotate counterclockwise, and the other two rotors M2 and M4 rotate clockwise. Roll is obtained by varying the speed of rotors M2 and M4 while the speed of rotors M1 and M3 remains constant

Similarly, pitch (θ) is achieved by varying the speed of rotors M1 and M3 while the speeds of rotors M2 and M4 remain the same. Yaw (ψ) is obtained by increasing (decreasing) the speed of rotors M1 and M3 and decreasing (increasing) the speed of rotors M2 and M4.

S1.2: Set various variable parameters to describe its position and motion attitude, and express the quadrotor flight motion with a simplified dynamic equation. In step 2, the controller structure in Fig. 3 is designed according to the dynamic equations and control parameters described below.

The specific method of step S1.2 is as follows:

The dynamic equation is used to express the relationship between the position, attitude, speed and acceleration of the quadrotor, and the preliminary modeling of the internal model of the quadrotor is realized. The modeling process is actually to obtain the relationship between force and acceleration. The simplified model of the position of the quadrotor in the case of external disturbance, that is, the translational motion equation is as follows:

分别表示沿x，y，z三轴的线加速度，m为四旋翼的质量，g为重力加速度，

θ，ψ分别表示滚转、俯仰、偏航的姿态角(即欧拉角)，表示机体框架相对于惯性框架的旋转，u1是四旋翼垂直z轴方向的控制力，

表示机体沿x，y，z三轴的线速度，k_x，k_y，k_z是x，y，z三轴的阻力系数，

表示外部干扰。In the formula,

Respectively represent the linear acceleration along the x, y, and z axes, m is the mass of the quadrotor, g is the acceleration of gravity,

θ and ψ respectively represent the attitude angles of roll, pitch and yaw (that is, the Euler angle), which represent the rotation of the body frame relative to the inertial frame, and u1 is the control force of the quadrotor in the vertical z-axis direction,

Indicates the linear velocity of the body along the three axes of x, y and z, k _x , k _y , k _z are the resistance coefficients of the three axes of x, y and z,

Indicates external interference.

The simplified model of the attitude of the quadrotor, that is, the rotation equation of motion is as follows:

分别表示滚转、俯仰、偏航的姿态角加速度，机体绕(x，y，z)三轴的角速度为

u₂，u₃，u₄分别为滚转、俯仰、偏航方向上的扭矩(即三个力矩的控制输出)，k₂，k₃，k₄是三个力矩的阻力系数，I_x，I_y，I_z分别为机体绕x，y，z轴的转动惯量，I_r为单个电机的转动惯量，Ω_r＝Ω₁-Ω₂+Ω₃-Ω₄为整体的转子角速度，Ω₁，Ω₂，Ω₃，Ω₄分别表示四个转子的角速度，

表示外部干扰。In the formula,

Represent the attitude angular acceleration of roll, pitch, and yaw respectively, and the angular velocity of the body around the three axes (x, y, z) is

u ₂ , u ₃ , u ₄ are the torques in the directions of roll, pitch, and yaw respectively (that is, the control output of the three moments), k ₂ , k ₃ , and k ₄ are the drag coefficients of the three moments, I _x , I _y , I _z are the moments of inertia of the body around the x, y, and z axes respectively, and I _r is the moment of inertia of a single motor, Ω _r =Ω ₁ -Ω ₂ +Ω ₃ -Ω ₄ is the overall rotor angular velocity, Ω ₁ , Ω ₂ , Ω ₃ , Ω ₄ represent the angular velocity of the four rotors respectively,

Indicates external interference.

S1.3 For the failure of the motor or propeller in the actuator, perform equation representation and description, and model related failures. The main content includes the operating principle of the actuator under normal and fault conditions and two types of fault models of the actuator.

The specific method of step S1.3 is as follows:

1) Working principle of actuator dynamics

The dynamic equations of each controller output, actuator control input and rotor speed of the quadrotor are as follows:

u ₁ =F ₁ +F ₂ +F ₃ +F ₄ (7)

u ₂ =(F ₄ -F ₂ )L (8)

u ₃ =(F ₃ -F ₁ )L (9)

u ₄ ＝(τ ₁ -τ ₂ +τ ₃ -τ ₄ ) (10)

和

表示第n个电机的产生的扭矩和推力，w_n为四个螺旋桨的转速，n＝1，2，3，4，b和d为常数，b为推力因子，d为阻力因子；L为四旋翼中心到各螺旋桨中心的距离。In the formula, u ₁ is the control force of the quadrotor in the direction of the vertical z-axis (that is, the total thrust), and u ₂ , u ₃ and u ₄ are the torques in the directions of roll, pitch and yaw, respectively (that is, roll angle control, pitch angle control, yaw angle control);

and

Indicates the torque and thrust produced by the nth motor, w _n is the rotational speed of the four propellers, n=1, 2, 3, 4, b and d are constants, b is the thrust factor, d is the resistance factor; L is four The distance from the center of the rotor to the center of each propeller.

2) Fault model of the actuator: as shown in Figure 2, the fault signal output of the actuator is independent of the input signal, where u(t) represents the control input [u ₁ , u ₂ , u ₃ , u ₄ ] at time t, y(t) represents the actual signal output (the final control signal output after the fault influence). Considering the actuator fault that occurs during the flight of the quadrotor, in order to express the impact of the fault on the system, a bounded additive fault u _f is added to the control output of the quadrotor and τ, 0≤τ≤1 is used to describe the actuator fault index. Different types of actuator faults can be described according to different values of τ and u _f . The actuator fault model equation is as follows:

y(t)=u _f +τu(t) (11)

where u _f represents a vector of four actuator additive fault values:

u _f =[f ₁ , f ₂ , f ₃ , f ₄ ] (12)

In the formula, f ₁ , f ₂ , f ₃ , and f ₄ represent the unknown constant additive faults of the four actuators.

2. Hybrid adaptive fault-tolerant control strategy

S2.1: Design of control structure based on double closed-loop control loop. Through the flight control model in step 1, it can be found that the quadrotor position motion and attitude motion have certain decoupling characteristics. Therefore, the quadrotor controller is divided into two subsystems for design, namely the altitude controller and the attitude controller. Given the position and attitude, the autonomous flight of the quadrotor can be realized. Secondly, a cascaded closed-loop control scheme is also used in the quadrotor control structure. After receiving the input command, the closed-loop controller will calculate the required change amount according to the feedback, and then adjust according to the embedded control algorithm, and correct the output to the actuator. The change of the output will also indirectly affect the feedback value, so the controller will continue to adjust until the system reaches the expected state. When the deviation between the UAV target state and the expected state is eliminated, the autonomous control task can be completed.

The specific method of step S2.1 is as follows:

u_θ分别控制，传感器测量值作为反馈提供给两个控制器。As shown in Fig. 3, the present invention designs a four-rotor cascaded double closed-loop control structure according to the equation and parameters in step S1.2, including two loop controls: an outer loop and an inner loop. The outer ring is the position ring, which mainly realizes the control of the position, acceleration, and speed of the UAV, tracks the position error and minimizes it, and provides the required attitude angle; the inner ring is the attitude ring, which is used to adjust the attitude angle of the aircraft to realize The control of the attitude angle of the quadrotor UAV. The outer loop controller generates thrust to control the altitude based on the desired altitude, and also generates the desired pitch and desired roll angles based on the difference between the desired horizontal position and the actual position. The inner loop controller inputs are the desired pitch, roll and yaw angles, and a motor control signal for the desired torque is generated based on the errors between the three desired and measured angles. In Fig. 3, u ₁ controls the command to generate thrust and is responsible for controlling the altitude, while u _ψ is responsible for controlling the yaw motion. The position controller provides the attitude controller with the required roll and pitch angles generated by

u _θ are controlled separately, and the sensor measurements are provided as feedback to the two controllers.

S2.2: Hybrid Adaptive Fault Tolerant Controller Based on Dual Closed Loops. The fault-tolerant control design of the system is mainly based on the double closed loop structure proposed in S2.1, including the position outer loop PID control structure and the attitude inner loop hybrid control structure, that is, the flight control structure adopts the external position control based on PID algorithm and the control structure based on The internal adaptive attitude control of the RL algorithm performs fault-tolerant control. This design combines a learning-based data-driven approach with a classical model PID control algorithm to form a hybrid adaptive fault-tolerant controller structure.

The specific method of step S2.2 is as follows:

A hybrid fault-tolerant control framework based on model controllers and reinforcement learning methods is proposed for actuator faults. The proposed control structure of the fault-tolerant control system is shown in Fig. 4, which implements a nested control structure to achieve complete control of the four rotors. The designed control strategy is based on two loops (inner loop and outer loop). The inner loop contains three control laws: roll control, pitch control, and yaw control. The outer loop includes three control laws for position x, y, z. The proposed flight control system stabilizes quadrotor autonomous flight despite the presence of UAV actuator failures.

1) Position loop PID control structure. The position control adopts the model-based PID algorithm, which retains the model-based robust method to ensure the stability and safety of the control.

The PID controller provides fast control response, which is the key to safe quadcopter flight. In addition, the PID controller has a simple structure, high stability and robustness. The PID controller is controlled by setting the error between the expected value and the actual value, and mainly includes three elements, namely proportional (K _p ), integral (K _I ) and derivative (K _D ), as shown in Figure 5. The core of the PID controller is to reduce the error Error to zero through the PID equation of formula (13). Among them, "error" is the difference between the set point (desired value) and the feedback amount (actual value).

where t represents time, x(t) is the bottom-level control output of the quadrotor, K _p , K _I and K _D are the proportional, integral and differential gain control parameters, respectively, e(t) is the set value (desired) and feedback ( Actual) difference between.

2) Attitude inner loop hybrid control structure. Attitude control adopts RL adaptive algorithm to adjust and update PID control parameters, which can design a more robust fault-tolerant controller for dynamic systems under fault conditions.

Although the model-based PID controller can provide fast control response, if an unexpected type of failure occurs, the actuator often cannot operate as expected, and the fault tolerance is insufficient. However, the response speed of the control algorithm based entirely on RL is slow, and cannot provide any analytical guarantee for the stable response and robustness of the control process. So the design of attitude fault-tolerant control in the present invention mixes RL algorithm with traditional PID control algorithm to provide an adaptive and robust fault-tolerant controller. To compensate the dynamic fluctuations of the aerial robot, in the proposed control structure, a reinforcement learning algorithm interacts with the system and learns an adaptive strategy to actively update and adjust the control parameters and weights of the attitude PID controller. The dynamic learning controller combining RL and PID can automatically optimize the parameters of the control law, avoiding the cost of manual parameter adjustment, and can adapt to changing environmental conditions or new faults by adjusting the weights of multiple model controllers, reducing environmental dynamic changes and interference, model failure effects.

In the present invention, the RL algorithm employs Deep Deterministic Policy Gradient (DDPG). DDPG is a method that combines the algorithmic ideas of DPG with deep Q-learning network (DQN). The DDPG algorithm is a hybrid technique that combines policy gradients and value functions. The architecture of DDPG is based on the actor-critic framework. The goal of the DDPG algorithm is to learn a parameterized deterministic policy μ _θ (s) such that the obtained optimal policy maximizes the expected reward over all states attainable by the policy.

其中，π表示策略函数，

分别表示第i个PID控制器对应的比例、积分和导数参数，

θ_d，

分别表示设置的期望俯仰角、期望滚转角和期望偏航角，最终的a(t)输出用于主动估计和调整内环多个PID控制器的增益系数和权重信息。critic网络提供用于训练更新actor信息。DDPG算法的网络架构如图6所示。The adaptive DDPG-PID attitude control method has a hierarchical structure as shown in Figure 4. In this structure, the RL algorithm plays the role of the upper controller, which can adjust the PID parameters (K _p , _KI , K _D ) and the participation weight δ ⁱ of the lower controller in real time according to the current state. The underlying PID controller compensates the error e(t) in the control process by outputting x(t). The action output of the actor network in the DDPG algorithm is the controller parameters and weight information, namely

Among them, π represents the policy function,

respectively represent the proportional, integral and derivative parameters corresponding to the i-th PID controller,

θ _d ,

Denote the set expected pitch angle, expected roll angle and expected yaw angle respectively, and the final a(t) output is used to actively estimate and adjust the gain coefficient and weight information of multiple PID controllers in the inner loop. The critic network provides actor information for training updates. The network architecture of the DDPG algorithm is shown in Figure 6.

In another embodiment, the present invention proposes an actuator failure-oriented quadrotor hybrid fault-tolerant control system corresponding to the actuator failure-oriented quadrotor hybrid fault-tolerant control method described in the first embodiment, including:

The calculation unit is used to establish the dynamic model of the quadrotor aircraft based on the principle of quadrotor flight, establish the ground reference coordinate system and the body coordinate system to represent the quadrotor flight structure, and use dynamic equations to represent the quadrotor flight motion and actuator failure model;

The control unit is used to design the flight control structure based on the double closed-loop control loop according to the quadrotor flight motion and the actuator failure model, and to carry out the fault-tolerant control design for the flight control structure, using the position outer loop control based on the PID method and the RL algorithm The self-adaptive attitude inner loop control realizes the quadrotor hybrid fault-tolerant control.

In another embodiment, the present invention provides a computer-readable storage medium, which stores a computer program, wherein the computer program enables the computer to execute the quadrotor oriented actuator failure as described in the first embodiment. Hybrid fault-tolerant control method.

In the embodiments disclosed in this application, a computer storage medium may be a tangible medium, which may contain or store a program for use by or in combination with an instruction execution system, device or device. Computer storage media may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or devices, or any suitable combination of the foregoing. More specific examples of computer storage media would include one or more wire-based electrical connections, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM), or flash memory), optical fiber, compact disc read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.

Those skilled in the art can appreciate that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed in the present application can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be regarded as exceeding the scope of the present application.

The above are only preferred implementations of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions under the idea of the present invention belong to the protection scope of the present invention. It should be pointed out that for those skilled in the art, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims (10)

1. A four-rotor hybrid fault-tolerant control method for actuator failure, characterized in that, comprising: Based on the principle of quadrotor flight, establish the dynamic model of the quadrotor aircraft, establish the ground reference coordinate system and the body coordinate system to represent the quadrotor flight structure, and use the dynamic equation to represent the quadrotor flight motion and actuator failure model; According to the quadrotor flight motion and actuator failure model, the flight control structure is designed based on the double closed-loop control loop, and the fault-tolerant control design is carried out for the flight control structure, and the position outer loop control based on the PID method and the adaptive attitude inner loop based on the RL algorithm are adopted. The loop control realizes the quadrotor hybrid fault-tolerant control. 2. a kind of four-rotor hybrid fault-tolerant control method facing actuator failure as claimed in claim 1, is characterized in that: described ground reference coordinate system and body coordinate system are respectively {R}(O, x, y, z ) and {R _b }(O _b , x _b , y _b , z _b ), where the ground reference coordinate system belongs to the Cartesian coordinate system and is fixed on the earth, the coordinate origin O is any point on the ground, and O _b is the center of mass of the quadrotor , {R _b } and {R} pass the quadrotor's position vector p=[x, y, z] ^T and the attitude angle vector

Associated. 3. A kind of quadrotor hybrid fault-tolerant control method facing actuator fault as claimed in claim 2, is characterized in that: described use dynamics equation to represent quadrotor flight motion and is specifically as follows: The translational motion equation of the quadrotor in the case of external disturbance is as follows:

In the formula,

Respectively represent the linear acceleration along the x, y, and z axes, m is the mass of the quadrotor, g is the acceleration of gravity,

θ, ψ represent the attitude angles of roll, pitch, and yaw respectively, and u ₁ is the control force of the quadrotor in the direction of the vertical z-axis,

Indicates the linear velocity of the body along the three axes of x, y and z, k _x , k _y , k _z are the resistance coefficients of the three axes of x, y and z,

Indicates external interference; The rotational motion equation of the quadrotor is as follows:

In the formula,

Represent the attitude angular acceleration of roll, pitch, and yaw respectively, and the angular velocity of the body around the three axes (x, y, z) is

u ₂ , u ₃ , u ₄ are the torques in the direction of roll, pitch, and yaw respectively; k ₂ , k ₃ , k ₄ are the drag coefficients of the three moments; I _x , I _y , I _z are the body’s winding The moment of inertia of the x, y, and z axes, I _r is the moment of inertia of a single motor, Ω _r =Ω ₁ -Ω ₂ +Ω ₃ -Ω ₄ is the overall rotor angular velocity, Ω ₁ , Ω ₂ , Ω ₃ , Ω ₄ represent the angular velocities of the four rotors respectively,

Indicates external interference. 4. a kind of actuator fault-oriented hybrid fault-tolerant control method for quadrotors as claimed in claim 2, is characterized in that: the establishment process of the actuator fault model is specifically as follows: The dynamic equations for establishing the output of each controller of the quadrotor, the control input of the actuator and the rotor speed are as follows: u ₁ =F ₁ +F ₂ +F ₃ +F ₄ (7) u ₂ =(F ₄ -F ₂ )L (8) u ₃ =(F ₃ -F ₁ )L (9) u ₄ ＝(τ ₁ -τ ₂ +τ ₃ -τ ₄ ) (10) In the formula, u ₁ is the control force of the quadrotor in the vertical z-axis direction, u ₂ , u ₃ and u ₄ are the torques in the roll, pitch and yaw directions respectively;

and

Indicates the torque and thrust generated by the nth motor, w _n is the speed of the four propellers, n=1, 2, 3, 4, b is the thrust factor, d is the resistance factor; L is the center of the four rotors to the center of each propeller distance; The actuator fault model equation is as follows: y(t)=u _f +τu(t) (11) Among them, u(t) represents the control input [u ₁ , u ₂ , u ₃ , u ₄ ] at time t, τ, 0≤τ≤1 represents the actuator fault index, and u _f represents the additive fault value of four actuators The resulting vector: u _f =[f ₁ , f ₂ , f ₃ , f ₄ ] (12) In the formula, f ₁ , f ₂ , f ₃ , and f ₄ represent the unknown constant additive faults of the four actuators. 5. A kind of four-rotor hybrid fault-tolerant control method facing actuator failure as claimed in claim 1, wherein: the flight control structure includes an outer loop controller, an inner loop controller and sensors, wherein the outer loop is the position The ring is used to control the position, acceleration and speed of the UAV, track the position error and minimize it, and generate the required attitude angle; the inner ring is the attitude ring, which is used to adjust the attitude angle of the aircraft; The outer loop controller generates the thrust to control the altitude according to the desired altitude, generates the expected pitch angle and the expected roll angle according to the difference between the expected position and the actual position and provides them to the inner loop controller; the inner loop controller inputs the expected pitch angle , the expected roll angle and the expected yaw angle, the expected yaw angle is calculated according to the expected position, the inner loop controller generates the motor control signal of the expected torque based on the error between the three expected angles and the measured angle; the sensor uses For the state measurement of the quadrotor aircraft, the measured value is provided as feedback to the outer loop controller and the inner loop controller. 6. A kind of four-rotor hybrid fault-tolerant control method facing actuator failure as claimed in claim 5, is characterized in that: the position outer loop control based on the PID method adopts the PID controller based on the dynamic model, through the PID equation To reduce the error to zero, the error is the difference between the expected value and the actual value, the PID equation is as follows:

In the formula, x(t) is the bottom control output of the quadrotor, t represents the time, K _p , K _I and K _D are the proportional, integral and differential gain control parameters respectively, and e(t) is the distance between the expected value and the actual value difference; The RL algorithm-based adaptive attitude inner-loop control adopts an RL adaptive algorithm to adjust and update the control parameters and weights of the PID controller. 7. a kind of four-rotor hybrid fault-tolerant control method facing actuator failure as claimed in claim 6, is characterized in that: described RL algorithm is as upper controller, adjusts the PID parameter (K of lower floor PID controller in real time according to current state) _p , K _I , K _D ) and participation weight δ ⁱ , the lower layer PID controller compensates the error e(t) generated in the control process by outputting x(t). 8. A kind of four-rotor hybrid fault-tolerant control method facing actuator failure as claimed in claim 7, is characterized in that: described RL self-adaptive algorithm adopts deep deterministic policy gradient DDPG, and the action output a of actor network in DDPG algorithm (t) is the controller parameters and weight information:

Among them, π represents the policy function,

respectively represent the proportional, integral and derivative parameters corresponding to the i-th PID controller,

Respectively represent the expected pitch angle, expected roll angle and expected yaw angle of the settings; the critic network provides actor information for training and updating. 9. A quadrotor hybrid fault-tolerant control system facing actuator faults, characterized in that it comprises: The calculation unit is used to establish the dynamic model of the quadrotor aircraft based on the principle of quadrotor flight, establish the ground reference coordinate system and the body coordinate system to represent the quadrotor flight structure, and use dynamic equations to represent the quadrotor flight motion and actuator failure model; The control unit is used to design the flight control structure based on the double closed-loop control loop according to the quadrotor flight motion and the actuator failure model, and to carry out the fault-tolerant control design for the flight control structure, using the position outer loop control based on the PID method and the RL algorithm The self-adaptive attitude inner loop control realizes the quadrotor hybrid fault-tolerant control. 10. A computer-readable storage medium storing a computer program, wherein the computer program causes the computer to execute the four-rotor hybrid fault-tolerant control method oriented to actuator failure according to any one of claims 1-8.

Images