Distributed propulsion aircraft power yaw control method based on deep reinforcement learning

Abstract

The application belongs to the technical field of flight control, and discloses a distributed propulsion aircraft power yaw control method based on deep reinforcement learning. The power yaw control method comprises the following steps: installing a distributed propeller; establishing a six-degree-of-freedom nonlinear dynamics model; obtaining aerodynamic data and pressure distribution under differential power; establishing a distributed propulsion aircraft power yaw control framework; and establishing a distributed propulsion aircraft power yaw control law. The power yaw control method introduces the expected index into the reward function of the algorithm, obtains a power yaw coupling control model through offline training and online use, solves the flight-propulsion-control coupling control problem of the distributed propulsion aircraft, and has important significance for ensuring the aircraft attitude stability of the distributed propulsion aircraft during power yaw, improving the heading stability and landing safety.

Description

Technical Field

[0001] This invention belongs to the field of flight control technology, specifically relating to a method for dynamic yaw control of distributed propulsion aircraft based on deep reinforcement learning. Background Technology

[0002] Distributed propulsion aircraft utilize multiple thrusters arranged on the aircraft's symmetrical plane to achieve flight control and power output. Each thruster possesses a degree of independence and autonomy, allowing for attitude and heading control by adjusting the thrust of each thruster and the deflection angle of the control surfaces. Compared to conventionally configured aircraft, distributed propulsion aircraft offer significant advantages in dealing with crosswinds and gusts. Particularly during landing in strong crosswinds, where the yaw moment provided by the rudder alone is insufficient to correct the course, distributed propulsion aircraft can enhance directional stability and improve landing safety through power differential.

[0003] Existing distributed propulsion aircraft dynamic yaw control technology mainly utilizes modern control theory to design dynamic yaw control laws for the aircraft. For example, the inventions disclosed in the Chinese Patent Document Database are titled "A Distributed Electric Propulsion Aircraft and Its Control Method" (CN114476093A) and "A Cooperative Control Method for Multiple Thrusters Applicable to Distributed Electric Propulsion Aircraft" (CN114137839A), as well as the nonlinear dynamic inverse control of dynamic yaw of distributed electric propulsion aircraft published in the Journal of Aerospace Power (https: / / doi.org / 10.13224 / j.cnki.jasp.20220222) [J / OL] (You Shun, Kou Peng, Yao Xuanyu, Wang Jing, Liang Deliang, Liang Zhe).

[0004] The above methods can calculate the required power of each thruster based on heading and altitude commands, study thrust power distribution strategies, and achieve stable heading flight under gust conditions. However, powered yaw will change the airflow field on the upper surfaces of both wings. The airflow velocity on the upper surface of the wing with slower engine speed is relatively slower, while the airflow velocity on the upper surface of the wing with faster engine speed is relatively faster. The side with faster airflow provides greater lift, which will generate additional roll moment, resulting in an unwanted roll angle on the aircraft. Especially during the landing phase in crosswind conditions, the additional roll angle may cause serious flight accidents such as wing touchdown and landing roll. In addition, when performing missions such as transporting personnel or surveying, the additional fuselage roll will affect passenger comfort and surveying accuracy. In this case, differential aileron adjustment is required to compensate for the roll moment generated by powered yaw. Different flight speeds, angles of attack, tail deflection angles, and power differential modes require different aileron differential compensation angles, which involve complex coupling relationships. Relying on traditional iterative fitting methods would consume a huge amount of manpower and financial resources, and would also have problems such as low mesh accuracy.

[0005] Currently, there is an urgent need to develop a distributed propulsion aircraft dynamic yaw control method based on deep reinforcement learning. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method for dynamic yaw control of distributed propulsion aircraft based on deep reinforcement learning.

[0007] The present invention provides a distributed propulsion aircraft dynamic yaw control method based on deep reinforcement learning, comprising the following steps:

[0008] S10. Install distributed thrusters;

[0009] The distributed propulsion aircraft has symmetrical wings, ailerons and V-tail on its fuselage; distributed propulsion units are installed on the upper surfaces of the wings on both sides, and the distributed propulsion units consist of several propulsion units symmetrically distributed on both sides.

[0010] S20. Establish a six-degree-of-freedom nonlinear dynamic model;

[0011] A six-degree-of-freedom nonlinear dynamic model was established based on the physical parameter relationships of the distributed thruster and the six-component aerodynamic data obtained from wind tunnel tests.

[0012] S30. Obtain aerodynamic data and pressure distribution under differential power;

[0013] Based on the pre-set differential mode of the two wing thrusters of the aircraft, wind tunnel tests or numerical simulations are conducted to obtain aerodynamic data and pressure distribution under the differential mode, and roll moment curves and yaw moment curves under differential thruster control are obtained.

[0014] S40. Establish a dynamic yaw control framework for distributed propulsion aircraft;

[0015] A dynamic differential control neural network was constructed and combined with the basic control law to obtain a distributed propulsion aircraft dynamic yaw control framework based on deep reinforcement learning;

[0016] S50. Establish a dynamic yaw control law for distributed propulsion aircraft;

[0017] A reward function is established, training parameters are provided, and a neural network is trained using Python to obtain a distributed propulsion aircraft dynamic yaw control law based on deep reinforcement learning.

[0018] Furthermore, the propulsion device is a ducted fan.

[0019] Furthermore, the physical parameter relationship of the distributed thruster is shown in the following formula:

[0020] ;

[0021] in, For flight speed, Sideslip angle, For the angle of attack, For the roll rate, For pitch rate, The yaw rate, For roll angle, The pitch angle, Yaw angle; For the rate of change of velocity, The sideslip angular velocity, The rate of change of angle of attack, The rate of change of roll angle, The rate of change of pitch angle, The rate of change of yaw angle. The rate of change of roll angle, This represents the rate of change of yaw angle; For quality, For dynamic pressure, For wing area, For wingspan, For the wing chord, It is the acceleration due to gravity; To bypass x Moment of inertia of the axis, To bypass y Moment of inertia of the axis, To bypass z Moment of inertia of the axis, The product of inertia; the six-component aerodynamic data represent the lift coefficient. Residual thrust coefficient Residual thrust coefficient The difference between the drag coefficient and the thrust coefficient, and the lateral force coefficient. Pitch moment coefficient Rolling torque coefficient and yaw moment coefficient .

[0022] Furthermore, the aforementioned six-degree-of-freedom nonlinear dynamic model is shown in the following equation:

[0023] ;

[0024] in, This represents the horizontal distance between the i-th thruster and the longitudinal axis of the spacecraft. This represents the remaining thrust coefficient of the i-th thruster; For aileron deflection, For elevator deflection angle, This refers to the rudder deflection angle; The lift coefficient at zero angle of attack. This represents the rate of change of the lift coefficient with respect to the angle of attack. This represents the rate of change of the lift coefficient with respect to the elevator deflection angle. The residual thrust coefficient at zero angle of attack. This represents the rate of change of the residual thrust coefficient with respect to the angle of attack. This represents the rate of change of the residual thrust coefficient with respect to the elevator deflection angle. The rate of change of the remaining thrust coefficient with respect to the square of the elevator deflection angle; This represents the rate of change of the lateral force coefficient with respect to the sideslip angle. This represents the rate of change of the lateral force coefficient with respect to the aileron deflection angle. This represents the rate of change of the lateral force coefficient with respect to the rudder deflection angle. This represents the rate of change of the rolling moment coefficient with respect to the rolling angular rate. This represents the rate of change of the roll moment coefficient with respect to the yaw rate. This represents the rate of change of the rolling moment coefficient with respect to the aileron deflection angle. This represents the rate of change of the roll moment coefficient with respect to the rudder deflection angle. This represents the rate of change of the rolling moment coefficient with respect to the sideslip angle. The initial velocity for the aircraft to perform a pitch maneuver. The pitch moment coefficient is the rate of change of velocity. This represents the rate of change of the pitch moment coefficient with respect to the angle of attack. This represents the rate of change of the pitch moment coefficient with respect to the pitch angular velocity. This represents the rate of change of the pitch moment coefficient with respect to the elevator deflection angle; This represents the rate of change of the yaw moment coefficient with respect to the sideslip angle. This represents the rate of change of the yaw moment coefficient with respect to the aileron deflection angle. This represents the rate of change of the yaw moment coefficient with respect to the rudder deflection angle. This represents the rate of change of the yaw moment coefficient with respect to the roll rate. This represents the rate of change of the yaw moment coefficient with respect to the yaw rate.

[0025] Furthermore, the input layer of the aforementioned dynamic differential control neural network includes an angle of attack. Roll angle Sideslip angle Aileron differential deflection and V-shaped tail wing deflection The neural network passes the parameters of the input layer to the hidden layer, and after calculation, obtains the output layer. , ... Choose the largest Q value, i.e. The optimal differential mode was obtained. For the selected dynamic differential control strategy;

[0026] in, For state Execution strategy The obtained Q value, For state, The subscripts 1, 2, ..., n represent different dynamic differential control strategies.

[0027] Furthermore, the basic control law includes:

[0028] Speed ​​control law:

[0029] ;

[0030] in, For speed tracking error, For flight speed reference signal; For speed proportional control parameters, These are speed integral control parameters; T For ducted fan power, This refers to the ducted fan's balance power during cruise mode.

[0031] Pitch angle control law:

[0032] ;

[0033] in, For pitch angle tracking error, This is the pitch angle reference signal; The pitch rate; For pitch angle control parameters, These are the pitch rate control parameters;

[0034] Roll angle control law:

[0035] ;

[0036] in, For roll angle tracking error, This is the roll angle reference signal; It is the roll rate; For roll angle control parameters, For roll rate control parameters;

[0037] To avoid roll angle If the angle is too large, design a roll angle protection system:

[0038] ;

[0039] in, This refers to the aileron deflection angle after amplitude limiting;

[0040] Heading angle control law:

[0041] ;

[0042] in, For sideslip angle tracking error, This is a sideslip angle reference signal; Yaw rate; This is the proportional control parameter for the sideslip angle. These are the integral control parameters for the sideslip angle. These are the yaw rate control parameters.

[0043] Furthermore, the reward function is:

[0044] ;

[0045] in, This is the reward value.

[0046] Furthermore, the power yaw control law for the distributed propulsion aircraft is as follows: First, the ducted fan power is calculated using the basic control law. T Elevator deflection Aileron deflection rudder deflection Then, the dynamic differential control neural network calculates the optimal power distribution strategy based on the input layer parameters.

[0047] Furthermore, the training parameters include decay coefficient, learning law, number of neural network layers and nodes, experience pool size, random sample size, discount factor, time step, and target network update speed.

[0048] The distributed propulsion yaw control method for aircraft based on deep reinforcement learning of the present invention has the following characteristics:

[0049] (1) It can calculate the distributed thruster power distribution strategy based on wind field information and stabilize the flight heading;

[0050] (2) It can provide aileron differential strategy based on information such as distributed thruster power distribution strategy, flight speed, and angle of attack to compensate for the additional roll caused by power differential and stabilize the attitude of distributed thruster;

[0051] (3) By adopting an offline training and online usage method, the time-consuming and computationally-intensive work is completed in the ground preparation work. The trained strategy network is directly written into the flight control board. Existing commercially available flight control boards can meet the computational power required for the flight of the distributed thruster, which has practical engineering value.

[0052] The present invention provides a deep reinforcement learning-based method for dynamic yaw control of distributed propulsion aircraft. This method incorporates the desired index into the reward function of the algorithm and obtains a coupled dynamic yaw control model through offline training and online application. This method solves the problem of flight-thrust-control coupled control of distributed propulsion aircraft and is of great significance for ensuring the attitude stability of the aircraft during dynamic yaw, improving heading stability and landing safety. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the structure of a distributed-powered aircraft.

[0054] Figure 2a The curve showing the effect of the dynamic differential on the yaw moment of a distributed-powered aircraft;

[0055] Figure 2b The curve showing the effect of dynamic differential on the rolling moment of a distributed-powered aircraft;

[0056] Figure 3 This is a diagram of the neural network for dynamic differential control.

[0057] Figure 4 This is a flowchart of the distributed propulsion aircraft dynamic yaw control method based on deep reinforcement learning according to the present invention.

[0058] In the diagram, 1. Thruster. Detailed Implementation

[0059] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0060] Example: The distributed powered aircraft of this example is shown below. Figure 1 , Figure 1 In the table, 1 indicates a thruster. The training parameters for this embodiment are shown in Table 1.

[0061] Table 1

[0062]

[0063] The distributed propulsion yaw control method for aircraft based on deep reinforcement learning in this embodiment includes the following steps:

[0064] S10. Install distributed thrusters;

[0065] The distributed propulsion aircraft has symmetrical wings, ailerons and V-tail on its fuselage; distributed propulsion units are installed on the upper surfaces of the wings on both sides, and the distributed propulsion units consist of several propulsion units symmetrically distributed on both sides.

[0066] S20. Establish a six-degree-of-freedom nonlinear dynamic model;

[0067] A six-degree-of-freedom nonlinear dynamic model was established based on the physical parameter relationships of the distributed thruster and the six-component aerodynamic data obtained from wind tunnel tests.

[0068] S30. Obtain aerodynamic data and pressure distribution under differential power;

[0069] Based on the pre-defined differential mode of the propulsion systems on both sides of the aircraft, wind tunnel tests or numerical simulations are conducted to obtain aerodynamic data and pressure distribution under the differential mode, resulting in... Figure 2a , Figure 2b The shown curves are the roll moment and yaw moment curves under differential control of the thruster.

[0070] S40. Establish a dynamic yaw control framework for distributed propulsion aircraft;

[0071] Building such Figure 3 The dynamic differential control neural network shown, combined with the basic control law, yields a distributed propulsion aircraft dynamic yaw control framework based on deep reinforcement learning;

[0072] S50. Establish a dynamic yaw control law for distributed propulsion aircraft;

[0073] Establish a reward function, provide training parameters, train a neural network using Python, and obtain the following results: Figure 4 The diagram shows a distributed propulsion aircraft dynamic yaw control law based on deep reinforcement learning.

[0074] Furthermore, the propulsion device is a ducted fan.

[0075] Furthermore, the physical parameter relationship of the distributed thruster is shown in the following formula:

[0076] ;

[0077] in, For flight speed, Sideslip angle, For the angle of attack, For the roll rate, For pitch rate, The yaw rate, For roll angle, The pitch angle, Yaw angle; For the rate of change of velocity, The sideslip angular velocity, The rate of change of angle of attack, The rate of change of roll angle, The rate of change of pitch angle, The rate of change of yaw angle. The rate of change of roll angle, This represents the rate of change of yaw angle; For quality, For dynamic pressure, For wing area, For wingspan, For the wing chord, It is the acceleration due to gravity; To bypass x Moment of inertia of the axis, To bypass y Moment of inertia of the axis, To bypass z Moment of inertia of the axis, The product of inertia; the six-component aerodynamic data represent the lift coefficient. Residual thrust coefficient Residual thrust coefficient The difference between the drag coefficient and the thrust coefficient, and the lateral force coefficient. Pitch moment coefficient Rolling torque coefficient and yaw moment coefficient .

[0078] Furthermore, the aforementioned six-degree-of-freedom nonlinear dynamic model is shown in the following equation:

[0079] ;

[0080] in, This represents the horizontal distance between the i-th thruster and the longitudinal axis of the spacecraft. This represents the remaining thrust coefficient of the i-th thruster; For aileron deflection, For elevator deflection angle, This refers to the rudder deflection angle; The lift coefficient at zero angle of attack. This represents the rate of change of the lift coefficient with respect to the angle of attack. This represents the rate of change of the lift coefficient with respect to the elevator deflection angle. The residual thrust coefficient at zero angle of attack. This represents the rate of change of the residual thrust coefficient with respect to the angle of attack. This represents the rate of change of the residual thrust coefficient with respect to the elevator deflection angle. The rate of change of the remaining thrust coefficient with respect to the square of the elevator deflection angle; This represents the rate of change of the lateral force coefficient with respect to the sideslip angle. This represents the rate of change of the lateral force coefficient with respect to the aileron deflection angle. This represents the rate of change of the lateral force coefficient with respect to the rudder deflection angle. This represents the rate of change of the rolling moment coefficient with respect to the rolling angular rate. This represents the rate of change of the roll moment coefficient with respect to the yaw rate. This represents the rate of change of the rolling moment coefficient with respect to the aileron deflection angle. This represents the rate of change of the roll moment coefficient with respect to the rudder deflection angle. This represents the rate of change of the rolling moment coefficient with respect to the sideslip angle. The initial velocity for the aircraft to perform a pitch maneuver. The pitch moment coefficient is the rate of change of velocity. This represents the rate of change of the pitch moment coefficient with respect to the angle of attack. This represents the rate of change of the pitch moment coefficient with respect to the pitch angular velocity. This represents the rate of change of the pitch moment coefficient with respect to the elevator deflection angle; This represents the rate of change of the yaw moment coefficient with respect to the sideslip angle. This represents the rate of change of the yaw moment coefficient with respect to the aileron deflection angle. This represents the rate of change of the yaw moment coefficient with respect to the rudder deflection angle. This represents the rate of change of the yaw moment coefficient with respect to the roll rate. This represents the rate of change of the yaw moment coefficient with respect to the yaw rate.

[0081] Furthermore, the input layer of the aforementioned dynamic differential control neural network includes an angle of attack. Roll angle Sideslip angle Aileron differential deflection and V-shaped tail wing deflection The neural network passes the parameters of the input layer to the hidden layer, and after calculation, obtains the output layer. , ... Choose the largest Q value, i.e. The optimal differential mode was obtained. The selected power differential control strategy;

[0082] in, For state Execution strategy The obtained Q value, For state, The subscripts 1, 2, ..., n represent different dynamic differential control strategies.

[0083] Furthermore, the basic control law includes:

[0084] Speed ​​control law:

[0085] ;

[0086] in, For speed tracking error, For flight speed reference signal; For speed proportional control parameters, These are speed integral control parameters; T For ducted fan power, This refers to the ducted fan's balance power during cruise mode.

[0087] Pitch angle control law:

[0088] ;

[0089] in, For pitch angle tracking error, This is the pitch angle reference signal; The pitch rate; For pitch angle control parameters, These are the pitch rate control parameters;

[0090] Roll angle control law:

[0091] ;

[0092] in, For roll angle tracking error, This is the roll angle reference signal; It is the roll rate; For roll angle control parameters, For roll rate control parameters;

[0093] To avoid roll angle If the angle is too large, design a roll angle protection system:

[0094] ;

[0095] in, This refers to the aileron deflection angle after amplitude limiting;

[0096] Heading angle control law:

[0097] ;

[0098] in, For sideslip angle tracking error, This is a sideslip angle reference signal; Yaw rate; This is the proportional control parameter for the sideslip angle. These are the integral control parameters for the sideslip angle. These are the yaw rate control parameters.

[0099] Furthermore, the reward function is:

[0100] ;

[0101] in, This is the reward value.

[0102] Furthermore, the power yaw control law for the distributed propulsion aircraft is as follows: First, the ducted fan power is calculated using the basic control law. T Elevator deflection Aileron deflection rudder deflection Then, the dynamic differential control neural network calculates the optimal power distribution strategy based on the input layer parameters.

[0103] Furthermore, the training parameters include decay coefficient, learning law, number of neural network layers and nodes, experience pool size, random sample size, discount factor, time step, and target network update speed.

[0104] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. For those skilled in the art, all features disclosed in the present invention, or all steps in all methods or processes disclosed, except for mutually exclusive features and / or steps, can be combined in any way without departing from the principles of the present invention. The present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for dynamic yaw control of a distributed propulsion aircraft based on deep reinforcement learning, characterized in that, The aforementioned dynamic yaw control method includes the following steps: S10. Install distributed thrusters; The distributed propulsion aircraft has symmetrical wings, ailerons and V-tail on its fuselage; distributed propulsion units are installed on the upper surfaces of the wings on both sides, and the distributed propulsion units consist of several propulsion units symmetrically distributed on both sides. S20. Establish a six-degree-of-freedom nonlinear dynamic model; A six-degree-of-freedom nonlinear dynamic model was established based on the physical parameter relationships of the distributed thruster and the six-component aerodynamic data obtained from wind tunnel tests. S30. Obtain aerodynamic data and pressure distribution under differential power; Based on the pre-set differential mode of the two wing thrusters of the aircraft, wind tunnel tests or numerical simulations are conducted to obtain aerodynamic data and pressure distribution under the differential mode, and roll moment curves and yaw moment curves under differential thruster control are obtained. S40. Establish a dynamic yaw control framework for distributed propulsion aircraft; A dynamic differential control neural network was constructed and combined with the basic control law to obtain a distributed propulsion aircraft dynamic yaw control framework based on deep reinforcement learning; The input layer of the aforementioned dynamic differential control neural network includes the angle of attack. Roll angle Sideslip angle Aileron differential deflection and V-shaped tail wing deflection The neural network passes the parameters of the input layer to the hidden layer, and after calculation, obtains the output layer. , ... Choose the largest Q value, i.e. The optimal differential mode was obtained. The selected dynamic differential control strategy; in, For state Execution strategy The obtained Q value, For state, The subscripts 1, 2, ..., n represent different dynamic differential control strategies; The basic control laws include: Speed ​​control law: ； in, For speed tracking error, For flight speed reference signal; For speed proportional control parameters, These are speed integral control parameters; T For ducted fan power, This refers to the ducted fan's balance power during cruise mode. Pitch angle control law: ； in, For pitch angle tracking error, This is the pitch angle reference signal; The pitch rate; For pitch angle control parameters, These are the pitch rate control parameters; Roll angle control law: ； in, For roll angle tracking error, This is the roll angle reference signal; It is the roll rate; For roll angle control parameters, For roll rate control parameters; To avoid roll angle If the angle is too large, design a roll angle protection system: ； in, This refers to the aileron deflection angle after amplitude limiting; Heading angle control law: ； in, For sideslip angle tracking error, This is a sideslip angle reference signal; Yaw rate; This is the proportional control parameter for the sideslip angle. These are the integral control parameters for the sideslip angle. These are the yaw rate control parameters; S50. Establish a dynamic yaw control law for distributed propulsion aircraft; A reward function is established, training parameters are provided, and a neural network is trained using Python to obtain a distributed propulsion aircraft dynamic yaw control law based on deep reinforcement learning.

2. The method for dynamic yaw control of a distributed propulsion aircraft based on deep reinforcement learning according to claim 1, characterized in that, The aforementioned propulsion device is a ducted fan.

3. The method for dynamic yaw control of a distributed propulsion aircraft based on deep reinforcement learning according to claim 2, characterized in that, The physical parameter relationship of the distributed thruster is shown in the following formula: ； in, For flight speed, Sideslip angle, For the angle of attack, For the roll rate, For pitch rate, The yaw rate is... For roll angle, The pitch angle, Yaw angle; For the rate of change of velocity, The sideslip angular velocity, For the rate of change of angle of attack, The rate of change of roll angle, The rate of change of pitch angle, The rate of change of yaw angle. The rate of change of roll angle, This represents the rate of change of yaw angle; For quality, For dynamic pressure, For wing area, For wingspan, For the wing chord, It is the acceleration due to gravity; To bypass x Moment of inertia of the axis, To bypass y Moment of inertia of the axis, To bypass z Moment of inertia of the axis, The product of inertia; the six-component aerodynamic data represent the lift coefficient. Residual thrust coefficient Residual thrust coefficient The difference between the drag coefficient and the thrust coefficient, and the lateral force coefficient. Pitch moment coefficient Rolling torque coefficient and yaw moment coefficient .

4. The distributed propulsion yaw control method for aircraft based on deep reinforcement learning according to claim 3, characterized in that, The six-degree-of-freedom nonlinear dynamic model is shown in the following equation: ； in, This represents the horizontal distance between the i-th thruster and the longitudinal axis of the spacecraft. This represents the remaining thrust coefficient of the i-th thruster; For aileron deflection, For elevator deflection angle, This refers to the rudder deflection angle; The lift coefficient at zero angle of attack. This represents the rate of change of the lift coefficient with respect to the angle of attack. This represents the rate of change of the lift coefficient with respect to the elevator deflection angle; The residual thrust coefficient at zero angle of attack. This represents the rate of change of the residual thrust coefficient with respect to the angle of attack. This represents the rate of change of the residual thrust coefficient with respect to the elevator deflection angle. The rate of change of the remaining thrust coefficient with respect to the square of the elevator deflection angle; This represents the rate of change of the lateral force coefficient with respect to the sideslip angle. This represents the rate of change of the lateral force coefficient with respect to the aileron deflection angle. This represents the rate of change of the lateral force coefficient with respect to the rudder deflection angle. This represents the rate of change of the rolling moment coefficient with respect to the rolling angular rate. This represents the rate of change of the roll moment coefficient with respect to the yaw rate. This represents the rate of change of the rolling moment coefficient with respect to the aileron deflection angle. This represents the rate of change of the roll moment coefficient with respect to the rudder deflection angle. This represents the rate of change of the rolling moment coefficient with respect to the sideslip angle. The initial velocity for the aircraft to perform a pitch maneuver. The pitch moment coefficient is the rate of change of velocity. This represents the rate of change of the pitch moment coefficient with respect to the angle of attack. This represents the rate of change of the pitch moment coefficient with respect to the pitch angular velocity. This represents the rate of change of the pitch moment coefficient with respect to the elevator deflection angle; This represents the rate of change of the yaw moment coefficient with respect to the sideslip angle. This represents the rate of change of the yaw moment coefficient with respect to the aileron deflection angle. This represents the rate of change of the yaw moment coefficient with respect to the rudder deflection angle. This represents the rate of change of the yaw moment coefficient with respect to the roll rate. This represents the rate of change of the yaw moment coefficient with respect to the yaw rate.

5. The distributed propulsion yaw control method for aircraft based on deep reinforcement learning according to claim 4, characterized in that, The reward function is as follows: ； in, This is the reward value.

6. The method for dynamic yaw control of a distributed propulsion aircraft based on deep reinforcement learning according to claim 5, characterized in that, The power yaw control law for the distributed propulsion aircraft is as follows: First, the ducted fan power is calculated using the basic control law. T Elevator deflection Aileron deflection rudder deflection Then, the dynamic differential control neural network calculates the optimal power distribution strategy based on the input layer parameters.

7. The method for dynamic yaw control of a distributed propulsion aircraft based on deep reinforcement learning according to claim 6, characterized in that, The training parameters include decay coefficient, learning law, number of neural network layers and nodes, experience pool size, random sample size, discount factor, time step, and target network update speed.

Images