Autopilot and its decoupling control method based on radial basis function neural network

Abstract

The present invention discloses an adaptive fully decoupled autopilot and a control method based on radial basis function (RBF) neural network control. The system comprises a required overload receiving module for receiving required overload information transmitted by a guidance system in real time, an aircraft parameter measuring module for obtaining flight parameters of an aircraft in real time, and a decoupling control module for obtaining available rudder instructions, wherein a rudder instruction for control decoupling is obtained according to the required overload information and the flight parameters of the aircraft, a transitional rudder instruction is obtained in combination with the flight parameters of the aircraft, and an available rudder instruction is obtained in combination with the flight parameters of the aircraft, and the steering operation of the servo is controlled accordingly; wherein, when the decoupling control module is used for decoupling calculation, the state feedback matrix and feedforward compensation matrix involved are obtained in real time through a radial basis function neural network model and the current status of the aircraft, thereby further improving the control performance.

Description

Technical Field

[0001] The invention relates to a control system and method for a rotating aircraft, and in particular to an adaptive fully decoupled autopilot and a control method based on radial basis function neural network control. Background Art

[0002] Rotating aircraft can bring many benefits by using the spinning method, such as effectively reducing the impact of aircraft structural design deviations on aircraft trajectory, simplifying control system design, omitting roll control mechanisms, etc. However, while this type of aircraft brings many benefits, it also has some disadvantages. Since the aircraft will generate a large roll angular velocity after spinning, the aircraft will produce characteristics such as aerodynamic coupling, inertial coupling, and control coupling, which makes the pitch channel and yaw channel mutually coupled and cross-linked, which is not conducive to the precise control of the pitch and yaw channels of the aircraft. In addition, the hysteresis characteristics of the servo have a great influence on the control of the rotating aircraft. Therefore, it is necessary to consider the influence of the servo link on the aircraft control system when designing the autopilot. The traditional design method often assumes that the servo is a first-order inertial link. This method ignores the control coupling characteristics of the rotating aircraft, and the servo model is not accurate enough, which increases the risk of the designed autopilot. If the servo is identified as a second-order inertial link, the dynamic model of the rotating aircraft is an eighth-order system. The model is more complex and involves inversion operations during the solution process. The inversion of the eighth-order system is too difficult, and the aircraft's processing equipment is difficult to solve in time. These problems have brought more severe challenges to the precise and stable control of the rotating aircraft control system.

[0003] In the prior art, in the process of actually controlling a rotating aircraft, the influence of the above-mentioned coupling and the second-order inertia link are ignored, so there are certain deviations in the actual control process of the aircraft, and the final guidance and control effect still needs to be improved.

[0004] In addition, the control parameters of the existing decoupled autopilot vary with the dynamic coefficient of the rotating aircraft, so one or more sets of autopilot control parameters are not sufficient to meet the actual application requirements.

[0005] In order to solve the above problems, the inventors have conducted in-depth research on existing autopilots in the hope of designing an adaptive fully decoupled autopilot based on radial basis function neural network control and a decoupling control method thereof that can solve the above problems. Summary of the invention

[0006] In order to overcome the above problems, the inventors have conducted intensive research and designed an adaptive fully decoupled autopilot and control method based on radial basis function neural network control, the system comprising a required overload receiving module for receiving required overload information transmitted by a guidance system in real time, an aircraft parameter measuring module for obtaining the flight parameters of the aircraft in real time, and a decoupling control module for obtaining available rudder instructions, wherein a rudder instruction for control decoupling is obtained according to the required overload information and the flight parameters of the aircraft, and then a transitional rudder instruction is obtained in combination with the flight parameters of the aircraft, and then an available rudder instruction is obtained in combination with the flight parameters of the aircraft, and the steering operation of the servo is controlled accordingly, wherein, when the decoupling control module is used for decoupling calculation, the state feedback matrix and feedforward compensation matrix involved are obtained in real time through the radial basis function neural network model and the current status of the aircraft, thereby further improving the control performance, thereby completing the present invention.

[0007] Specifically, the purpose of the present invention is to provide an adaptive fully decoupled autopilot based on radial basis neural network control, the system is installed on a rotating aircraft, the system includes

[0008] The required overload receiving module 1 is connected to the guidance system on the rotating aircraft to receive the required overload information transmitted by the guidance system in real time.

[0009] Aircraft parameter measurement module 2, which is used to obtain the flight parameters of the aircraft in real time,

[0010] A decoupling control module 3 is used to obtain available rudder instructions in real time according to the required overload information and the flight parameters of the aircraft, and

[0011] The radial basis neural network model 4 is used for solving in real time the state feedback matrix and feedforward compensation matrix required for solving the available steering instructions.

[0012] The aircraft parameter measurement module 2 includes a steering gear attitude sensor 21, a steering gear angular rate sensor 22, an accelerometer 23, an inertial gyroscope 24 and an estimator 25;

[0013] The steering gear attitude sensor 21 is used to measure the pitch rudder deflection angle information and the yaw rudder deflection angle information of the aircraft in real time.

[0014] The servo angular rate sensor 22 is used to measure the pitch servo angular rate information and the yaw servo angular rate information of the aircraft in real time.

[0015] The accelerometer 23 is used to measure the acceleration information of the aircraft in real time.

[0016] The inertial gyro 24 is used to measure the yaw rate information and pitch rate information of the aircraft in real time.

[0017] The estimator 25 is configured to estimate the angle of attack information and sideslip angle information of the aircraft in real time based on the acceleration information.

[0018] Among them, the decoupling control module 3 includes a control decoupling rudder command resolver module 31, a transition rudder command resolver module 32, and an available rudder command resolver module 33;

[0019] Among them, the control decoupling rudder command resolver module 31 is configured to obtain the control decoupling rudder command in real time according to the required overload information and the flight parameters of the aircraft;

[0020] The transition rudder command resolver module 32 is configured to obtain the transition rudder command in real time according to the flight parameters of the aircraft and the control decoupling rudder command;

[0021] The available rudder command resolver module 33 is configured to obtain the available rudder command in real time according to the flight parameters of the aircraft and the transition rudder command.

[0022] Among them, the control decoupling rudder command resolver module 31 obtains the control decoupling rudder command in real time through the following formula (1),

[0023] u2 = -K2x2 + L2v2 (1)

[0024] Among them, u2 represents the control decoupling rudder command, K2 represents the state feedback matrix, L2 represents the feedforward compensation matrix, x2 represents the state variable of the actuator state space expression, and v2 represents the required overload.

[0025] Among them, the transition rudder command resolver module 32 obtains the transition rudder command in real time through the following formula (2),

[0026] y2 = C2∫(A2x2 + B2u2)dt (2)

[0027] Among them, y2 represents the transition rudder command,

[0028] A2, B2, and C2 all represent actuator system parameters.

[0029] Among them, the available rudder command resolver module 33 obtains the available rudder command in real time through the following formula (3),

[0030] u1 = -K1x1 + L1v1 (3)

[0031] Among them, u1 represents the available rudder command, K1 represents the state feedback matrix, L1 represents the feedforward compensation matrix, x1 represents the state variable of the aircraft state space expression, and v1 represents the transition rudder command.

[0032] The decoupling control module 3 is connected to a steering gear 5 of the aircraft, and the steering gear 5 steers according to available rudder instructions.

[0033] Before the rotating aircraft is launched, a certain number of characteristic points at different heights and speeds are selected as samples in the simulated aircraft trajectory to train the radial basis function neural network model 4;

[0034] Preferably, after training, the radial basis neural network model 4 can output the state feedback matrix K1, the state feedback matrix K2, the feedforward compensation matrix L1 and the feedforward compensation matrix L2 according to the expected pitch overload, the expected yaw overload, the rotating aircraft flight speed and the rotating aircraft flight altitude information in real time.

[0035] The present invention also provides a decoupling control method of an adaptive fully decoupling autopilot based on radial basis neural network control, the method comprising the following steps:

[0036] Step 1, receiving the required overload information transmitted by the guidance system through the overload receiving module 1;

[0037] Step 2, obtaining the flight parameters of the aircraft through the aircraft parameter measurement module 2;

[0038] Step 3, obtaining available rudder instructions according to the required overload information and the flight parameters of the aircraft through the decoupling control module 3;

[0039] Step 4, repeat steps 1-3 in real time, so as to obtain available rudder instructions in real time.

[0040] Wherein, the step 2 includes the following sub-steps:

[0041] Sub-step 2-1, obtaining the pitch rudder deflection angle information and the yaw rudder deflection angle information of the aircraft in real time through the servo attitude sensor 21,

[0042] Sub-step 2-2, obtaining the pitch servo angular rate information and the yaw servo angular rate information of the aircraft in real time through the servo angular rate sensor 22,

[0043] Sub-step 2-3, obtaining the acceleration information of the aircraft by real-time measurement through the accelerometer 23, and obtaining the yaw rate information and pitch rate information of the aircraft by real-time measurement through the inertial gyroscope 24,

[0044] Sub-step 2-4, the estimator 25 estimates the angle of attack and sideslip angle of the aircraft in real time based on the three-axis acceleration information.

[0045] Wherein, the step 3 includes the following sub-steps:

[0046] Sub-step 3-1, obtaining the control decoupling rudder instruction in real time according to the required overload information and the flight parameters of the aircraft through the control decoupling rudder instruction solver module 31;

[0047] Sub-step 3-2, obtaining the transitional rudder instruction in real time according to the flight parameters of the aircraft and the control decoupled rudder instruction through the transitional rudder instruction solving submodule 32;

[0048] Sub-step 3-3, obtaining available rudder instructions in real time according to the flight parameters of the aircraft and the transitional rudder instructions through the available rudder instruction solving submodule 33.

[0049] The beneficial effects of the present invention include:

[0050] (1) The adaptive fully decoupled autopilot based on radial basis function neural network control and its decoupling control method can provide more reasonable servo control instructions for the servo based on the received required overload, combined with the current flight status of the rotating aircraft, and considering the second-order servo dynamics and coupling conditions, thereby enhancing the control effect of the rotating aircraft and improving the control accuracy of the rotating aircraft;

[0051] (2) The adaptive fully decoupled autopilot based on radial basis neural network control and its decoupling control method take into account the second-order servo power and simplify the calculation process. The eighth-order system is transformed into two fourth-order systems by model transformation for solution control, which saves calculation time and can meet the requirements of real-time information acquisition of the aircraft.

[0052] (3) In the adaptive fully decoupled autopilot based on radial basis function neural network control and its decoupling control method, the control parameters of the autopilot, namely the feedforward compensation matrix and the state feedback matrix, are obtained in real time through neural network control, thereby further improving the accuracy of the decoupling control. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The overall logic diagram of an adaptive fully decoupled autopilot based on radial basis function neural network control according to a preferred embodiment of the present invention is shown;

[0054] Figure 2 The required overload and response curve in the pitch direction in the simulation experiment are shown;

[0055] Figure 3 The required overload and response curve in the yaw direction in the simulation experiment are shown;

[0056] Figure 4 The pitch angle rate variation curve in the simulation experiment is shown;

[0057] Figure 5 The yaw angle rate change curve in the simulation experiment is shown.

[0058] Description of Figure Numbers:

[0059] 1-Overload receiving module is required

[0060] 2- Aircraft parameter measurement module

[0061] 3- Decoupling control module

[0062] 21-Servo attitude sensor

[0063] 22-Servo angular rate sensor

[0064] 23-Accelerometer

[0065] 24-Inertial Gyroscope

[0066] 25-Estimator

[0067] 26-Satellite signal receiving module

[0068] 31-Control decoupled rudder command solver module

[0069] 32-Transitional rudder command solver module

[0070] 33- Available rudder command solver modules

[0071] 4-Radial Basis Neural Network Model

[0072] 5-Servo DETAILED DESCRIPTION

[0073] The present invention will be further described in detail below through the accompanying drawings and embodiments. Through these descriptions, the characteristics and advantages of the present invention will become more clear and distinct.

[0074] The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise noted.

[0075] According to the present invention, an adaptive fully decoupled autopilot based on radial basis function neural network control is provided. Figure 1As shown in, the system is installed on a rotating aircraft, and the rotating aircraft is preferably a high-dynamic rotating aircraft, that is, a rotating aircraft with a rotation speed of more than 10r / s; the coupling described in this application means that when the pitch direction and yaw direction of the aircraft are controlled separately, the control instruction in one direction will affect and interfere with the other direction, especially when the pitch direction of the aircraft is controlled, due to the rotation, the force generated by the steering gear is likely to have a certain component in the horizontal direction, and this component will cause the aircraft to deflect in the yaw direction.

[0076] The system comprises a required overload receiving module 1, an aircraft parameter measuring module 2 and a decoupling control module 3;

[0077] The overload receiving module 1 is connected to the guidance system on the rotating aircraft to receive the required overload information transmitted by the guidance system in real time.

[0078] The guidance system is also installed on the aircraft, and it can give the required overload in real time according to the aircraft's own information and target information obtained by the sensors on the aircraft. In general, the steering engine is controlled to steer according to the required overload. However, in the solution provided by the present application, the required overload is not directly transmitted to the steering engine, but is transmitted to the required overload receiving module 1, and the available rudder command obtained after processing is transmitted to the steering engine. This makes the steering of the steering engine more targeted and has a better control effect on the rolling aircraft.

[0079] The guidance system is an existing guidance system in the art, which may adopt existing guidance laws such as proportional guidance law, gravity compensation guidance law, etc.

[0080] The aircraft parameter measurement module 2 is used to obtain the flight parameters of the aircraft in real time, and the flight parameters include pitch rudder deflection angle, yaw rudder deflection angle, pitch servo angular rate, yaw servo angular rate, acceleration, speed, yaw angular rate, pitch angular rate, angle of attack and sideslip angle; the aircraft parameter measurement module 2 can retrieve the dynamic coefficients related to the flight parameters in real time from the storage chip it carries.

[0081] The decoupling control module 3 is used to obtain available rudder instructions in real time according to the required overload information and the flight parameters of the aircraft, and transmit the rudder instructions to the steering gear, which steers according to the rudder instructions.

[0082] In a preferred embodiment, the required overload information includes expected pitch overload and expected yaw overload; the expected pitch overload is the overload required to be provided in the pitch direction calculated by the guidance system; the expected yaw overload is the overload required to be provided in the yaw direction calculated by the guidance system.

[0083] The control decoupled rudder instructions, transitional rudder instructions and available rudder instructions are all rudder instructions, and the main difference lies in the precision and accuracy. The rudder instructions include pitch rudder instructions and yaw rudder instructions; the pitch rudder instruction represents the instruction that is ultimately to be transmitted to the servo and executed by the servo in the pitch direction; the yaw rudder instruction represents the instruction that is ultimately to be transmitted to the servo and executed by the servo in the yaw direction.

[0084] When the guidance system of the aircraft obtains the required overload, if the required overload is directly transmitted to the servo, the servo will inevitably solve the corresponding pitch rudder command according to the expected pitch overload, and solve the corresponding yaw rudder command according to the expected yaw overload. During the execution of the servo, the lag characteristics of the servo have a great influence on the control of the rotating aircraft, and the traditional design method often defaults to the servo as a first-order inertia link. This method ignores the control coupling characteristics of the rotating aircraft, resulting in a large deviation between the final steering result and the expected value. After the required overload is solved by the decoupling control module, the influence of interference factors such as servo lag and coupling is calculated in advance, so that the steering work is performed according to the finally available rudder command, which can make the steering result closer to the expected value and achieve better control effect.

[0085] In a preferred embodiment, the aircraft parameter measurement module 2 includes a steering gear attitude sensor 21, a steering gear angular rate sensor 22, an accelerometer 23, an inertial gyroscope 24 and an estimator 25;

[0086] The steering gear attitude sensor 21 is used to measure the pitch rudder deflection angle information and yaw rudder deflection angle information of the aircraft in real time;

[0087] The servo angular rate sensor 22 is used to measure the pitch servo angular rate information and the yaw servo angular rate information of the aircraft in real time;

[0088] The accelerometer 23 is used to measure the acceleration information of the aircraft in real time.

[0089] The inertial gyro 24 is used to measure the yaw rate information and pitch rate information of the aircraft in real time.

[0090] The estimator 25 is used to estimate the angle of attack and the sideslip angle of the aircraft in real time according to the three-axis acceleration information.

[0091] Among them, a plurality of accelerometers 23 are provided. Preferably, at least 3 are provided. At least one accelerometer is located at the center of mass of the aircraft, and its installation direction is along the axis of the aircraft towards the traveling direction of the aircraft, for measuring the acceleration of the aircraft along the axis direction, that is, the acceleration of the aircraft itself. After integration of this acceleration, the speed information of the aircraft can be obtained;

[0092] In addition, there are two other accelerometers both installed on the axis of the aircraft and deviated from the center of mass by a certain distance, and the installation directions of these two accelerometers are perpendicular to each other. These two accelerometers are both connected to the estimator 25. The accelerometer can measure the acceleration value at its own position in real time. After integration, the speed at this point can be obtained, and then multiplied by the distance between this point and the center of mass to obtain the angular velocity at this point, and then integrated to obtain the angle. Preferably, the estimator is also connected to the geomagnetic sensor on the aircraft. The geomagnetic sensor can know the roll angle of the aircraft in real time, so that the sideslip angle and angle of attack of the aircraft can be obtained respectively through these two accelerometers and the roll angle information. The distance between the accelerometer and the center of mass is stored in the estimator, and integral calculation can also be performed in the estimator, so that the estimator can give the sideslip angle information and angle of attack information of the aircraft in real time.

[0093] Preferably, the aircraft parameter measurement module 2 further includes a satellite signal receiving module 26, which is used to receive satellite signals in real time and calculate the flight speed and flight height of the rotating aircraft.

[0094] In a preferred embodiment, the decoupling control module 3 includes a control decoupling rudder command resolver module 31, a transition rudder command resolver module 32, and an available rudder command resolver module 33;

[0095] Among them, the control decoupling rudder command resolver module 31 is used to obtain the control decoupling rudder command according to the required overload information and the flight parameters of the aircraft in real time;

[0096] The transition rudder command resolver module 32 is used to obtain the transition rudder command according to the flight parameters of the aircraft and the control decoupling rudder command in real time;

[0097] The available rudder command resolver module 33 is used to obtain the available rudder command according to the flight parameters of the required aircraft and the transition rudder command in real time.

[0098] Preferably, the control decoupling rudder command resolver module 31 obtains the control decoupling rudder command in real time through the following formula (1),

[0099] u2 = -K2x2 + L2v2 (1)

[0100] Among them, u2 represents the rudder command of control decoupling, K2 represents the state feedback matrix, L2 represents the feedforward compensation matrix, x2 represents the state variable of the servo state space expression, and v2 represents the required overload;

[0101] Preferably,

[0102]

[0103] v2=[v y v z ] T

[0104] δ y represents the pitch rudder deflection angle, δ z represents the yaw rudder deflection angle, Indicates the pitch servo angular rate, Yaw servo angular rate; v y represents the expected pitch overload, v z Indicates the expected yaw overload;

[0105] Preferably, the transitional rudder instruction solver module 32 obtains the transitional rudder instruction in real time through the following formula (II):

[0106] y2=C2∫(A2x2+B2u2)dt (II)

[0107] Among them, y2 represents the transitional rudder command,

[0108] A2, B2, and C2 all represent the parameters of the servo system.

[0109]

[0110]

[0111]

[0112] d3, d 11 d 12 d 21 d 22 Both represent the dynamic coefficients of the servo model, which can be obtained in real time based on the aircraft's own parameters and the aircraft's own speed.

[0113] Preferably, the available rudder command solver module 33 obtains the available rudder command in real time through the following formula (III):

[0114] u1=-K1x1+L1v1 (III)

[0115] Among them, u1 represents the available rudder command, K1 represents the state feedback matrix, L1 represents the feedforward compensation matrix, x1 represents the state variable of the aircraft state space expression, and v1, like y2, represents the transitional rudder command;

[0116]

[0117] α represents the angle of attack, β represents the sideslip angle, represents the pitch angle rate, Indicates the yaw rate.

[0118] In a preferred embodiment, the dynamic coefficient of the steering gear model is obtained by the following formula:

[0119]

[0120] in, Indicates the speed, which is detected in real time, μ s Indicates the servo damping ratio, T s Indicates the delay of the servo command, k s Indicates the servo gain, μ s 、T s and k s These are all parameters pre-installed in the aircraft when it leaves the factory.

[0121] In a preferred embodiment, the decoupling control module 3 is connected to a steering engine 5 of the aircraft, and the steering engine 5 steers according to available rudder instructions;

[0122] The actual overload obtained by the steering gear 5 in accordance with the available rudder command can be obtained by the following formula (IV):

[0123] y1=C1∫(A1x1+B1u1)dt (IV)

[0124] Where y1 = [a y a z ] T , indicating that the actual response is overloaded, a y Indicates the pitch direction response overload, a z Indicates the yaw direction response overload. A1, B1, and C1 all represent aircraft system parameters.

[0125]

[0126]

[0127]

[0128] a 25 、a 24 、a 27、a 22 、a 28 and a 34 Both represent the dynamic coefficient of a rotating aircraft. The dynamic coefficient of the rotating aircraft is known data pre-installed in the aircraft. It is generally calculated through wind tunnel experiments and other methods before the aircraft leaves the factory. The data can be retrieved at any time during the flight of the aircraft.

[0129] In a preferred embodiment, the radial basis function neural network model 4 is pre-installed on the aircraft. Before the rotating aircraft is launched, a certain number of characteristic points at different heights and speeds are selected as sampling points in the simulated aircraft trajectory, and the radial basis function neural network model 4 is obtained through sample flushing training;

[0130] The heights of the characteristic points in the sampling points are different, and the flight speeds of the aircraft at the characteristic points are also different;

[0131] Preferably, 12 groups of feature points are selected, and 144 groups of feature points can be obtained by arranging and combining the 12 groups of feature points. Each group of feature points is substituted into the solution formula (I), (II), and (III) of the decoupling control module 3 to calculate the state feedback matrix K1, K2 and the feedforward compensation matrix L1, L2. A group of calculation results and the corresponding feature points are taken as a sample, and the radial basis neural network model is obtained by flushing training through the 144 groups of samples.

[0132] Preferably, after training, the radial basis neural network model 4 can output the state feedback matrix K1, the state feedback matrix K2, the feedforward compensation matrix L1 and the feedforward compensation matrix L2 according to the expected pitch overload, the expected yaw overload, the rotating aircraft flight speed and the rotating aircraft flight altitude information in real time.

[0133] The specific information of the 12 groups of feature points at different speeds at different heights is exemplified as follows:

[0134] The heights are 200m, 500m, 1000m, 2000m, 3000m, 4000m, 5000m, 6000m, 7000m, 8000m, 9000m, 10000m; the corresponding speeds are 685m / s, 880m / s, 1080m / s, 950m / s, 700m / s, 585m / s, 294m / s, 395m / s, 486m / s, 535m / s, 640m / s, 720m / s.

[0135] Due to the axial symmetry of the rotating aircraft, the designed control parameters also have symmetrical characteristics, and the control parameters of the feedforward compensation matrix L2 are only related to the expected natural frequency. Therefore, the control parameters that need to be trained can be simplified to the first row control parameters K of the state feedback matrix K1 11 , K12 , K 13 , K 14 , the first row of the state feedback matrix K2 controls the parameters K 21 , K 22 , K 23 , K 24 And the first row control parameter L of the feedforward compensation matrix L1 11 , L 12 , then 144 sets of the above control parameters need to be calculated before training the radial basis neural network model.

[0136] The following Tables 1, 2 and 3 respectively give the control parameters of the state feedback matrix K1, K2 and the feedforward compensation matrix L1 at different speeds when the altitude H is 8000m, and the following Tables 4, 5 and 6 respectively give the control parameters of the state feedback matrix K1, K2 and the feedforward compensation matrix L1 at different altitudes when the speed V is 535m / s.

[0137] Table 1

[0138]

[0139] Table 2

[0140]

[0141] Table 3

[0142]

[0143] Table 4

[0144]

[0145] Table 5

[0146]

[0147] Table 6

[0148]

[0149] The present invention also provides a decoupling control method for an adaptive fully decoupling autopilot based on radial basis function neural network control, the method is implemented by the adaptive fully decoupling autopilot based on radial basis function neural network control as described above,

[0150] The method comprises the following steps,

[0151] Step 1, receiving the required overload information transmitted by the guidance system through the overload receiving module 1;

[0152] Step 2, obtaining the flight parameters of the aircraft through the aircraft parameter measurement module 2;

[0153] Step 3, obtaining available rudder instructions according to the required overload information and the flight parameters of the aircraft through the decoupling control module 3;

[0154] Step 4, repeat steps 1-3 in real time, so as to obtain available rudder instructions in real time.

[0155] Preferably, step 2 includes the following sub-steps:

[0156] Sub-step 2-1, obtaining the pitch rudder deflection angle information and the yaw rudder deflection angle information of the aircraft in real time through the servo attitude sensor 21,

[0157] Sub-step 2-2, obtaining the pitch servo angular rate information and the yaw servo angular rate information of the aircraft in real time through the servo angular rate sensor 22,

[0158] Sub-step 2-3, obtaining the acceleration information of the aircraft by real-time measurement through the accelerometer 23, and obtaining the yaw rate information and pitch rate information of the aircraft by real-time measurement through the inertial gyroscope 24,

[0159] Sub-step 2-4, the estimator 25 estimates the angle of attack and sideslip angle of the aircraft in real time based on the three-axis acceleration information.

[0160] Preferably, the step 3 includes the following sub-steps:

[0161] Sub-step 3-1, obtaining the angle of attack convergence error and the sideslip angle convergence error in real time according to the required overload information and the flight parameters of the aircraft by controlling the decoupled rudder command solver submodule 31;

[0162] Sub-step 3-2, obtaining the control decoupling rudder instruction in real time according to the required overload information and the flight parameters of the aircraft through the control decoupling rudder instruction solver module 31;

[0163] Sub-step 3-2, obtaining the transitional rudder instruction in real time according to the flight parameters of the aircraft and the control decoupled rudder instruction through the transitional rudder instruction solving submodule 32;

[0164] Sub-step 3-3, obtaining available rudder instructions in real time according to the flight parameters of the aircraft and the transitional rudder instructions through the available rudder instruction solving submodule 33.

[0165] Before executing step 1, the radial basis function neural network model is installed in the aircraft, and the radial basis function neural network model is used to calculate the state feedback matrices K1, K2 and the feedforward compensation matrices L1, L2 in real time according to the speed and altitude information of the aircraft.

[0166] Experimental example:

[0167] Directly simulate the guidance system and servo system of a rotary aircraft through a computer. The guidance system can give guidance commands in real time, that is, the required overload, specifically including the desired pitch overload and the desired yaw overload. The desired pitch overload is in a sinusoidal alternating state, with an amplitude of 10 m / s² and a frequency of 1 rad / s. The yaw overload command is zero. The trajectory of the required overload changing with time is as shown by the solid line in Figure 2 and Figure 3 ; The servo system controls the servo to deflect according to the guidance command or the required overload, and directly gives the overload situation that can actually be provided for the rotary aircraft after the servo works according to the required overload;

[0168] In the experimental example, intercept the required overload given by the guidance system in the computer, and do not directly transfer the required overload to the servo system. Instead, transfer the required overload to the adaptive full decoupling autopilot (RBF) based on radial basis neural network control provided by this application and the decoupling autopilot (Gain-Schedule) using the gain scheduling method;

[0169] After the required overload is transferred to the RBF, the available rudder command is obtained through the decoupling control method of the adaptive full decoupling autopilot based on radial basis neural network control provided by this application, and then the available rudder command is transferred to the servo system to control the servo to work accordingly, and the overload situation that can actually be provided for the rotary aircraft after the servo works is obtained;

[0170] Among them, the required overload information is received through the required overload receiving module, and the flight parameters of the simulated aircraft are given in real time by the computer, including the pitch rudder deflection angle, yaw rudder deflection angle, pitch servo angular rate, yaw servo angular rate, angle of attack, sideslip angle, speed, pitch angle and yaw angle; And the dynamic coefficients of the rotary aircraft are given as follows:

[0171]

[0172] The servo parameter table is as follows:

[0173]

[0174] The pitch-direction servo response command and the yaw-direction servo response command are calculated and obtained through the following formulas (1), (2), and (3);

[0175] u2 = -K2x2 + L2v2 (1)

[0176] y2 = C2∫(A2x2 + B2u2)dt (2)

[0177] u1 = -K1x1 + L1v1 (3)

[0178] v2 = [v y vz ] T , v y represents the expected pitch overload, v z Indicates the expected yaw overload,

[0179]

[0180] α represents the angle of attack, β represents the sideslip angle, represents the pitch angle rate, represents the yaw rate, δ y represents the pitch rudder deflection angle, δ z represents the yaw rudder deflection angle, Indicates the pitch servo angular rate, Indicates the yaw servo angular rate; u1 indicates the available rudder command, including the pitch servo response command and the yaw servo response command;

[0181]

[0182]

[0183]

[0184] L1, L2, K1, and K2 are obtained by real-time solution of the base neural network model;

[0185] After the servo response command is transmitted to the servo system in the computer, the simulation obtains the actual overload conditions that the servo can provide to the rotating aircraft after working. Figure 2 , Figure 3 , Figure 4 and Figure 5 The dashed line “RBF” is shown in the figure.

[0186] In the comparative example, after the required overload is transmitted to the Gain-Schedule, the Gain-Schedule system responds to the required overload, obtains the steering gear control instruction, and controls the steering gear accordingly, and finally obtains the overload condition provided by the steering gear, such as Figure 2 , Figure 3 , Figure 4 and Figure 5 This is shown by the dotted line “Gain-Schedule” in the figure.

[0187] in, Figure 2 is the pitch overload response curve, Figure 3 is the yaw overload response curve, Figure 4 is the pitch angle rate change curve, Figure 5 is the yaw rate variation curve. Figure 2 and Figure 3It can be seen that the pitch overload response of the traditional gain scheduling method has obvious steady-state errors, and the yaw channel also has an overload response that fluctuates in the form of a sinusoidal wave. It can be seen that the system is not completely decoupled and there is still a coupling effect; while the pitch overload response of the adaptive fully decoupled autopilot based on radial basis function neural network control is basically consistent with the overload command, and the yaw channel has almost no coupling response, which can accurately track the required overload and achieve complete decoupling of the inertial coupling, aerodynamic coupling and control coupling of the rotating aircraft. Figure 4 and Figure 5 It can be seen that the pitch rate of the traditional gain scheduling method is small, and the yaw rate has large fluctuations, which leads to steady-state errors in pitch overload and coupling effects in the yaw channel. In contrast, the adaptive fully decoupled autopilot based on radial basis function neural network control has better control performance and decoupling performance than the gain scheduling method when the sinusoidal signal is input.

[0188] The present invention has been described above in conjunction with preferred embodiments, but these embodiments are only exemplary and serve only as an illustration. On this basis, the present invention may be subjected to a variety of substitutions and improvements, all of which fall within the scope of protection of the present invention.

Claims

1. An adaptive fully decoupled autopilot based on radial basis neural network control, characterized in that: The autopilot is installed on a rotary aircraft, and the autopilot includes: A required overload receiving module (1), which is connected to the guidance system on the rotary aircraft to receive the required overload information transmitted by the guidance system in real time, An aircraft parameter measurement module (2), which is used to obtain the flight parameters of the aircraft in real time, A decoupling control module (3), which is used to obtain available rudder commands in real time according to the required overload information and the flight parameters of the aircraft; and A radial basis neural network model (4), which is used to calculate the state feedback matrix and the feedforward compensation matrix required for calculating the available rudder commands in real time; The decoupling control module (3) is connected to the rudder actuator (5) of the aircraft, and the rudder actuator (5) steers according to the available rudder commands; The actual response overload obtained by the rudder actuator (5) steering according to the available rudder commands can be known through the following formula (IV): y1 = C1∫(A1x1 + B1u1)dt (IV) Where y1 = [a y a z ] T , indicating the actual response overload, a y Indicates the pitch direction response overload, a z Indicates the yaw direction response overload. A1, B1, and C1 all represent the aircraft system parameters. a 25 、a 24 、a 27 、a 22 、a 28 and a 34 All represent the dynamic coefficients of a rotating aircraft; The radial basis neural network model (4) is pre-installed on the aircraft. Before the rotary aircraft is launched, a certain number of characteristic points with different heights and different speeds are selected as sampling points in the simulated aircraft trajectory, and the radial basis neural network model (4) is obtained through sample flushing training; The decoupling control module (3) includes a rudder command decoupling operator module (31) for control decoupling, a rudder command decoupling operator module (32) for transition, and a rudder command decoupling operator module (33) for availability; Among them, the rudder command decoupling operator module (31) for control decoupling is used to obtain the rudder command for control decoupling in real time according to the required overload information and the flight parameters of the aircraft; The rudder command decoupling operator module (32) for transition is used to obtain the rudder command for transition in real time according to the flight parameters of the aircraft and the rudder command for control decoupling; The rudder command decoupling operator module (33) for availability is used to obtain the available rudder command in real time according to the flight parameters of the aircraft and the rudder command for transition; The rudder command decoupling operator module (31) for control decoupling obtains the rudder command for control decoupling in real time through the following formula (I), u2 = -K2x2 + L2v2 (I) Where, u2 represents the rudder command for control decoupling, K2 represents the state feedback matrix, L2 represents the feedforward compensation matrix, x2 represents the state variable of the rudder actuator state space expression, and v2 represents the required overload; The rudder command decoupling operator module (32) for transition obtains the rudder command for transition in real time through the following formula (II), y2 = C2∫(A2x2 + B2u2)dt (II) Where, y2 represents the rudder command for transition, A2, B2, and C2 all represent the parameters of the rudder actuator system; The rudder command decoupling operator module (33) for availability obtains the available rudder command in real time through the following formula (III), u1 = -K1x1 + L1v1 (III) Where, u1 represents the available rudder command, K1 represents the state feedback matrix, L1 represents the feedforward compensation matrix, x1 represents the state variable of the aircraft state space expression, and v1 represents the rudder command for transition; The decoupling control module (3) is connected to the rudder actuator (5) of the aircraft, and the rudder actuator (5) steers according to the available rudder commands.

2. The adaptive fully decoupled autopilot based on radial basis neural network control according to claim 1, characterized in that: The aircraft parameter measurement module (2) comprises a steering gear attitude sensor (21), a steering gear angular rate sensor (22), an accelerometer (23), an inertial gyroscope (24) and an estimator (25); The steering gear attitude sensor (21) is used to measure in real time the pitch rudder deflection angle information and the yaw rudder deflection angle information of the aircraft. The servo angular rate sensor (22) is used to measure in real time the pitch servo angular rate information and the yaw servo angular rate information of the aircraft. The accelerometer (23) is used to measure the acceleration information of the aircraft in real time. The inertial gyro (24) is used to measure in real time the yaw rate information and pitch rate information of the aircraft. The estimator (25) is used to estimate the angle of attack and sideslip angle information of the aircraft in real time according to the acceleration information.

3. The adaptive fully decoupled autopilot based on radial basis neural network control according to claim 1, characterized in that: Before the rotating aircraft is launched, a certain number of characteristic points at different heights and speeds are selected as sampling points in the simulated aircraft trajectory, and the radial basis function neural network model (4) is obtained through sample flushing training; After being trained, the radial basis neural network model (4) can output a state feedback matrix K1, a state feedback matrix K2, a feedforward compensation matrix L1 and a feedforward compensation matrix L2 in real time according to information on expected pitch overload, expected yaw overload, the flight speed of the rotating aircraft and the flight altitude of the rotating aircraft.

4. A decoupling control method for an adaptive fully decoupled autopilot based on radial basis neural network control, characterized in that: The method uses an autopilot as claimed in any one of claims 1 to 3; The method comprises the following steps, Step 1, receiving the required overload information transmitted by the guidance system through the required overload receiving module (1); Step 2, obtaining the flight parameters of the aircraft through the aircraft parameter measurement module (2); Step 3, obtaining available rudder instructions according to the required overload information and the flight parameters of the aircraft through the decoupling control module (3); Step 4, repeat steps 1-3 in real time, so as to obtain available rudder instructions in real time.

5. The decoupling control method of the adaptive fully decoupling autopilot based on radial basis neural network control according to claim 4 is characterized in that: The step 2 includes the following sub-steps: Sub-step 2-1, obtaining the pitch rudder deflection angle information and the yaw rudder deflection angle information of the aircraft by real-time measurement through the servo attitude sensor (21), Sub-step 2-2, obtaining the pitch servo angular rate information and the yaw servo angular rate information of the aircraft by real-time measurement through the servo angular rate sensor (22), Sub-step 2-3, obtaining the acceleration information of the aircraft by real-time measurement through the accelerometer (23), and obtaining the yaw rate information and pitch rate information of the aircraft by real-time measurement through the inertial gyroscope (24), Sub-step 2-4, using an estimator (25) to estimate in real time the angle of attack and sideslip angle of the aircraft based on the acceleration information.

6. The decoupling control method of the adaptive fully decoupling autopilot based on radial basis function neural network control according to claim 4, characterized in that: The step 3 includes the following sub-steps: Sub-step 3-1, obtaining a control decoupling rudder instruction in real time according to the required overload information and the flight parameters of the aircraft through a control decoupling rudder instruction solving submodule (31); Sub-step 3-2, obtaining a transitional rudder instruction in real time according to the flight parameters of the aircraft and the control decoupled rudder instruction through a transitional rudder instruction solving submodule (32); Sub-step 3-3, obtaining available rudder instructions in real time according to the flight parameters of the aircraft and the transitional rudder instructions through the available rudder instruction solving submodule (33).

Images