CN102289714B - Method for controlling autonomous take-off and landing of small unmanned rotorcraft based on behavioral model - Google Patents

Abstract

The invention provides a method for controlling autonomous take-off and landing of a small unmanned rotorcraft based on a behavioral model, relating to flight control operator behavioral data learning, flight control operator behavioral model building and autonomous take-off and landing rule formulation. The method comprises the following steps: firstly acquiring flight control operator control behavioral data by a data acquisition system; then aiming at the operating characteristics of the small unmanned rotorcraft in the autonomous take-off and landing stage, building a flight control operator behavioral model by using a fuzzy control method and optimizing the model by a neural network to improve the performances of the flight control operator behavioral model; and formulating the autonomous take-off and landing behavior rule by analyzing the flight control operator behavioral characteristics and the take-off and landing characteristics of the small unmanned rotorcraft. The method has the advantages of strong antijamming capability, high stability, convenience in design and the like and can be used for controlling autonomous take-off and landing of the small unmanned rotorcraft in a complex environment.

Description

The autonomous landing control method of a kind of miniature self-service giro based on behavior model

Technical field

The present invention relates to the autonomous landing control method of a kind of miniature self-service giro based on behavior model, be applicable to work in aerial unmanned robot autonomous control field.

Background technology

Miniature self-service giro has the characteristics such as vertical takeoff and landing, hovering, can execute the task at narrow spaces such as drives, is with a wide range of applications.Along with the expansion of application, the intelligent degree demand of miniature self-service giro also increases day by day, and full miniature self-service giro autonomous, high intelligence becomes the focus of research.

As complicated multi-input multi-output control system, miniature self-service giro has strong coupling, non-linear, control difficulty characteristic.At present, the landing of miniature self-service giro is mainly controlled by manual remote control mode of operation, depends on to a great extent the technical ability and the skill level that fly to control hand self.But in actual applications,, most of armies mouth and civilian's user of service does not have the technical ability of controlling of specialty, needs miniature self-service giro research and development department to train for a long time user, thereby has limited the promotion and application of miniature self-service giro.

For improving performance, PID control method, robust control, all kinds of control methods such as intelligent control method are used to the autonomous landing of miniature self-service giro and control.Intelligent PID controller is controlled unmanned gyroplane attitude, improves system control accuracy, but poor anti jamming capability, and miniature self-service giro landing stage ground effect is disturbed, wind is disturbed greatly, intelligent PID control method is difficult to realize stable autonomous landing function.The perturbation that when unmanned vehicle flight is eliminated in robust control, department pattern parameter occurs, but there is the characteristic of real-time, dynamic parameter low-response.The nonlinear adaptive of neural network is controlled, and overcomes uncertainty, model-free and the potential kinematic nonlinearity of depopulated helicopter parameter, realizes the attitude of depopulated helicopter and controls, but need a large amount of sample calculation.

Summary of the invention

Technology of the present invention is dealt with problems and is: not enough for the existing control method of miniature self-service giro, the autonomous landing control method of a kind of miniature self-service giro based on behavior model is proposed, and solve miniature self-service giro and independently played drop controller design problem.

Technical solution of the present invention is: the present invention proposes a kind of autonomous landing control method based on Self-adaptive flight hand behavior model, collection flies to control hand behavioral data, analysis of small unmanned gyroplane plays drop characteristic, determine rule of conduct, method based on flying to control the study of hand behavioral data, structure flies to control hand behavior model, and by radial base neural net, flies to control hand behavior model and optimize, and concrete steps are as follows:

(1) gather and fly to control hand behavioral data

Flying to control hand controls miniature self-service giro and carries out landing operation, by data acquisition system (DAS), gather the data of controlling of the total distance of main slurry in miniature self-service giro landing stage and throttle threshold values, and the corresponding horizontal level of miniature self-service giro, highly, speed state information.

(2) build and fly to control hand behavior model

By study, fly to control hand and control behavioral data and miniature self-service giro status information, utilize adaptive fuzzy control method to build and fly to control hand behavior model; This model consists of input layer, obscuring layer, rules layer and deblurring layer, represents IF-THEN (condition conclusion) control procedure;

Input layer output node $O_i^{\left(1\right)}=X_i,i=\mathrm{1,2}---\left(1\right)$

Wherein, X _ifor the input of input layer, be respectively height error Δ height and height error variable quantity

output for input layer;

Obscuring layer input node $I_{\mathrm{ik}}^{\left(2\right)}=-{(X_i-a_{\mathrm{ik}})}^2/(k_i\&CenterDot;{{\mathrm{<figure-callout\; id="2"\; label="b"\; filenames="HDA0000065776690000041.png"\; state="\{\{state\}\}">b</figure-callout>}}^2}_{\mathrm{ik}}),i=1,2,k=\mathrm{1,2}\&CenterDot;\&CenterDot;\&CenterDot;N---\left(2\right)$

Output node $O_{\mathrm{ik}}^{\left(2\right)}=\exp \left(I_{\mathrm{ik}}^{\left(2\right)}\right)---\left(3\right)$

Wherein, a _ikrepresent that i inputs the central value of corresponding k class Gaussian function,

represent that i inputs the mean square deviation of corresponding k class Gaussian function, k _ifor the regulation and control parameter of radial base neural net to i input, N is fuzzy variable number corresponding to input;

Rules layer generates corresponding N based on MAX-MIN (minimax) product synthetic method ²individual fuzzy rule, it inputs node

$I_l^{\left(3\right)}=o_{1m}^{\left(2\right)}\&cap;o_{2n}^{\left(2\right)}=\exp \left(I_{1m}^{\left(2\right)}\right)\&cap;\exp \left(I_{2n}^{\left(2\right)}\right)$

l＝N·(m-1)+n m＝1，2…N，n＝1，2…N (4)

Output node $O_l^{\left(3\right)}=I_l^{\left(3\right)}---\left(5\right)$

Deblurring layer input node $I^{\left(4\right)}=\underset{j=1}{\overset{\mathrm{ll}}{\mathrm{\&Sigma;}}}{o}_j^{\left(3\right)}\&CenterDot;w_j,\mathrm{ll}=N^2---\left(6\right)$

Output node $O^{\left(4\right)}=k_3\&CenterDot;I^{\left(4\right)}/\underset{j=1}{\overset{\mathrm{ll}}{\mathrm{\&Sigma;}}}{o}_j^{\left(3\right)},\mathrm{ll}=N^2---\left(7\right)$

Wherein, w _jthe corresponding weights of j rule, k ₃for the regulation and control parameter of radial base neural net to i input, O ⁽⁴⁾for always apart from the controlled quentity controlled variable of rudder amount.

Select 150 groups of autonomous landing training results of miniature self-service giro based on flying to control hand behavior model under different bumpy weathers as sample, build radial base neural net and come real-time optimization to fly to control the Gaussian function mean square deviation of hand behavior model obscuring layer and always apart from the controlled quentity controlled variable of rudder amount; Radial base neural net is three-layer neural network, comprises input layer, and three layers of radial base neural net that hidden layer and output layer form form; Input layer output node is

${\mathrm{OO}}_i^{\left(1\right)}=X{X}_i,i=\mathrm{1,2}---\left(8\right)$

Wherein, XX _ifor the input of input layer, be respectively horizontal level difference DELTA lateral and the speed difference DELTA speed of liftoff stage of miniature self-service giro when floating state,

output for input layer;

Hidden layer input node: $I_c^{\left(2\right)}=\underset{i=1}{\overset{2}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ic}}{\mathrm{XX}}_i,c=\mathrm{1,2}\&CenterDot;\&CenterDot;\&CenterDot;Z---\left(9\right)$

Output node: ${\mathrm{OO}}_c^{\left(2\right)}=\exp ({(I_c^{\left(2\right)}-{\mathrm{aa}}_c)}^2/{\mathrm{bb}}_c^2),c=\mathrm{1,2}\&CenterDot;\&CenterDot;\&CenterDot;Z---\left(10\right)$

Wherein, aa _cthe central value that represents c corresponding Gaussian function,

the mean square deviation that represents c corresponding Gaussian function, Z is hidden layer neuron number;

Output layer input node ${I_p}^{\left(3\right)}=\underset{e=1}{\overset{Z}{\mathrm{\&Sigma;}}}{\mathrm{oo}}_{\mathrm{pe}}^{\left(2\right)}\&CenterDot;w_{\mathrm{pe}},p=\mathrm{1,2,3}---\left(11\right)$

Output node k _p ⁽³⁾=I _p ⁽³⁾(12)

The output k of output layer wherein ₁for for adjusting the coefficient of Gaussian function corresponding to height error, k ₂for adjusting the coefficient of Gaussian function corresponding to height error, k ₃for adjusting output always apart from the coefficient of rudder amount.

(3) design autonomous landing rule of conduct

Initial period is by throttle and always apart from threshold values, obtain adhesion, the liftoff stage flies to control hand behavior model by self-adaptation and eliminates the lengthwise position height that external disturbance obtains and control, PID carries out position, speed and the attitude of transverse plane and controls, and utilizes PID to carry out position, speed and attitude control at mission phase; The Autonomous landing stage is divided into mission phase, liftoff stage and landing stage.Mission phase is realized attitude by PID and is controlled, and the liftoff stage flies to control hand behavior model implementation lengthwise position height by self-adaptation and controls, and PID realizes horizontal level, speed and attitude and controls, and the stage that lands is realized position control by flying to control hand behavior model.

The present invention's advantage is compared with prior art:

(1) the present invention learns to fly to control hand by adaptive fuzzy control method and controls behavioral data, simulation skillfully flies to control hand and controls behavior, a little less than the dependence of miniature self-service giro model, and can be according to the real flight conditions real-time online principle of optimality, antijamming capability is strong;

(2) the present invention is based on train fly to control hand behavior model, in conjunction with traditional control method, can according to the status information of aircraft, adjust rudder amount in real time, calculated amount is little, real-time is good, dynamic parameter fast response time;

(3) the present invention proposes flies to control hand behavior model based on fuzzy control method, study flies to control hand behavioral data, and carry out performance real-time optimization by radial basis function network, on reasonably regular basis, only need less sample just can draw rational rule.

Accompanying drawing explanation

Fig. 1 is the autonomous control flow of miniature self-service giro;

Fig. 2 is that miniature self-service giro self-adaptation flies to control hand behavior model;

Fig. 3 is the miniature self-service giro Three-dimensional Track figure that independently takes off;

Fig. 4 is miniature self-service giro Autonomous landing Three-dimensional Track figure.

Embodiment

As shown in Figure 1, 2, concrete methods of realizing of the present invention is as follows:

(1) gather and fly to control hand behavioral data

Flying to control hand controls miniature self-service giro and carries out landing operation, by data acquisition system (DAS) according to time clock with the 10Hz sampling period, gather and fly to control hand and control input data at the total distance of main slurry in miniature self-service giro landing stage and throttle control, according to miniature self-service giro state of flight determine the threshold values that takes off of total distance and throttle, the control threshold value of hovering phase, landing the stage idling threshold value, with horizontal position information, the height status information of miniature self-service giro in the landing stage, velocity information builds and flies to control hand behavior model training sample.

(2) build and fly to control hand behavior model

The miniature self-service giro height error collecting by learning data acquisition system and height error variable quantity status information and corresponding main slurry, always apart from input quantity, utilize adaptive fuzzy control method to build and fly to control hand behavior model; This model consists of input layer, obscuring layer, rules layer and deblurring layer, represents IF-THEN control procedure;

Input layer output node $O_i^{\left(1\right)}=X_i,i=\mathrm{1,2}---\left(1\right)$

Wherein, X _ifor the input of input layer, be respectively true altitude and the height error Δ height between Desired Height and height error variable quantity that data acquisition system (DAS) collects

output for input layer;

Obscuring layer input node $I_{\mathrm{ik}}^{\left(2\right)}=-{(X_i-a_{\mathrm{ik}})}^2/(k_i\&CenterDot;{{\mathrm{<figure-callout\; id="2"\; label="b"\; filenames="HDA0000065776690000041.png"\; state="\{\{state\}\}">b</figure-callout>}}^2}_{\mathrm{ik}}),i=1,2,k=\mathrm{1,2}\&CenterDot;\&CenterDot;\&CenterDot;N---\left(2\right)$

Output node $O_{\mathrm{ik}}^{\left(2\right)}=\exp \left(I_{\mathrm{ik}}^{\left(2\right)}\right)---\left(3\right)$

Wherein, a _ikrepresent that i inputs the central value of corresponding k class Gaussian function, represent that i inputs the mean square deviation of corresponding k class Gaussian function, k _ifor the regulation and control parameter of radial base neural net to i input, N is fuzzy variable number corresponding to input;

Rules layer generates corresponding N based on MAX-MIN (minimax) product synthetic method ²individual fuzzy rule, it inputs node

$I_l^{\left(3\right)}=o_{1m}^{\left(2\right)}\&cap;o_{2n}^{\left(2\right)}=\exp \left(I_{1m}^{\left(2\right)}\right)\&cap;\exp \left(I_{2n}^{\left(2\right)}\right)$

l＝N·(m-1)+n m＝1，2…N，n＝1，2…N (4)

Output node $O_l^{\left(3\right)}=I_l^{\left(3\right)}---\left(5\right)$

Deblurring layer input node $I^{\left(4\right)}=\underset{j=1}{\overset{\mathrm{ll}}{\mathrm{\&Sigma;}}}{o}_j^{\left(3\right)}\&CenterDot;w_j,\mathrm{ll}=N^2---\left(6\right)$

Output node $O^{\left(4\right)}=k_3\&CenterDot;I^{\left(4\right)}/\underset{j=1}{\overset{\mathrm{ll}}{\mathrm{\&Sigma;}}}{o}_j^{\left(3\right)},\mathrm{ll}=N^2---\left(7\right)$

Wherein, w _jthe corresponding weights of j rule, k ₃for the regulation and control parameter of radial base neural net to i input, O ⁽⁴⁾for always apart from the controlled quentity controlled variable of rudder amount, a _ik,

and w _jby genetic algorithm, train and obtain.

Select 150 groups of autonomous landing training results of miniature self-service giro based on flying to control hand behavior model under different bumpy weathers as sample, build radial base neural net and come real-time optimization to fly to control the Gaussian function mean square deviation of hand behavior model obscuring layer and always apart from the controlled quentity controlled variable of rudder amount; Radial base neural net is three-layer neural network, comprises input layer, and three layers of radial base neural net that hidden layer and output layer form form; Input layer output node is

${\mathrm{OO}}_i^{\left(1\right)}=X{X}_i,i=\mathrm{1,2}---\left(8\right)$

Wherein, XX _ifor the input of input layer, be respectively the horizontal level in liftoff stage of miniature self-service giro and the horizontal level difference DELTA lateral of desired locations and speed difference DELTA speed that data acquisition system (DAS) collects,

output for input layer;

Hidden layer input node: $I_c^{\left(2\right)}=\underset{i=1}{\overset{2}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ic}}{\mathrm{XX}}_i,c=\mathrm{1,2}\&CenterDot;\&CenterDot;\&CenterDot;Z---\left(9\right)$

Output node: ${\mathrm{OO}}_c^{\left(2\right)}=\exp ({(I_c^{\left(2\right)}-{\mathrm{aa}}_c)}^2/{\mathrm{bb}}_c^2),c=\mathrm{1,2}\&CenterDot;\&CenterDot;\&CenterDot;Z---\left(10\right)$

Wherein, aa _cthe central value that represents c corresponding Gaussian function,

the mean square deviation that represents c corresponding Gaussian function, Z is hidden layer neuron number;

Output layer input node ${I_p}^{\left(3\right)}=\underset{e=1}{\overset{Z}{\mathrm{\&Sigma;}}}{\mathrm{oo}}_{\mathrm{pe}}^{\left(2\right)}\&CenterDot;w_{\mathrm{pe}},\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="HDA0000065776690000041.png"\; state="\{\{state\}\}">\; <figure-callout\; id="2"\; label="p"\; filenames="HDA0000065776690000041.png"\; state="\{\{state\}\}">p</figure-callout>\; </figure-callout>}\mathrm{1,2,3}---\left(11\right)$

Output node k _p ⁽³⁾=I _p ⁽³⁾(12)

The output k of output layer wherein ₁for for adjusting the coefficient of Gaussian function corresponding to height error, k ₂for adjusting the coefficient of Gaussian function corresponding to height error, k ₃for adjusting output always apart from the coefficient of rudder amount.

(3) design autonomous landing rule of conduct

Analysis of small unmanned gyroplane flight characteristics takeoff phase, will mainly be divided into initial period, liftoff stage and mission phase autonomous takeoff phase.At initial period, miniature self-service giro receives and independently takes off after instruction, using current as takeoff point, by to flying to control the study of hand behavioral data, the threshold value of taking off that at the uniform velocity increases throttle and always collect apart from controlled quentity controlled variable to data acquisition system (DAS), to obtaining expectation rotating speed and lift, when reaching throttle and total distance, throttle carries out saturated control, and miniature self-service giro enters into liftoff critical conditions; The liftoff stage flies to control hand behavior model acquisition lengthwise position height by self-adaptation and controls, PID method is carried out pitching and the rolling of transverse plane and is controlled, the course error of resolving based on magnetic compass is carried out attitude locking control, when barometric leveling is greater than after 4m, enters mission phase; At mission phase, using takeoff point overhead 10m as target hover point, the coupling by pitching, rolling, course and height loop controls, and realizes autonomous hovering and controls.

Analysis of small unmanned gyroplane landing stage flight characteristics, is mainly divided into mission phase, liftoff stage and landing stage by the Autonomous landing stage.At mission phase, miniature self-service giro receives after landing instruction, and based on current state, take 10 meters, overhead, level point is target hover point, and the coupling by pitching, rolling, course and height loop controls, and realizes autonomous hovering and controls; Then based on horizontal position error, Negotiation speed and attitude realize stable control, the course error of resolving based on magnetic compass is carried out attitude locking and is controlled, based on pressure information, obtain true altitude and fall high control information with expectation, according to the pattern of the little amplitude limit of vast scale, high control independently fallen, when measuring height is less than after 2 meters, enter the liftoff stage; In the liftoff stage, position and attitude that miniature self-service giro is realized surface level by pid loop are controlled, and fly to control hand behavior model realize the height of fore-and-aft plane and control by self-adaptation, when measuring height is less than after 0.2 meter, enter the landing stage; Take horizontal attitude as expectation attitude angle, based on stance loop, realize pitching, rolling, jaw channel control, by flying to control hand behavior model, to take the landing stage threshold values that data acquisition system (DAS) collects total apart from locking as expectation value enters, and control throttle and in 4 seconds, be at the uniform velocity reduced to the throttle at idle that data acquisition system (DAS) collects the miniature self-service giro landing stage, reduce lift, realize stable landing, surpass after 4 seconds, system is directly carried out definite value control by total distance and throttle, obtain larger adhesion, improve system reliability.

Flight example

Based on thunder tiger 90 SUAV (small unmanned aerial vehicle), carry out the autonomous landing control method of the miniature self-service giro of flight validation based on behavior model, miniature self-service giro can be realized stable autonomous landing function; As shown in Figure 3, miniature self-service giro receives and independently takes off after instruction autonomous flight track, with (7,-3,0), as takeoff point, complete aerial mission in 10 seconds, final hover point site error is less than 2m, and height error is less than 1m, and hovering velocity error is less than 0.5m/s; Miniature self-service rotor craft Autonomous landing process as shown in Figure 4, miniature self-service giro receives after landing instruction in arbitrfary point, directly autonomous target hover point (0,0,10), in 15 seconds, complete autonomous landing process, due to ground effect function influence, miniature self-service giro actual falling point is that (1,2,0) site error is less than 2m.

The autonomous landing control method of miniature self-service giro that the present invention is based on behavior model has overcome the deficiency of existing control method, can realize complete autonomous landing control under miniature self-service giro complex environment etc.

The content not being described in detail in instructions of the present invention belongs to the known prior art of professional and technical personnel in the field.

Claims (2)

1. the autonomous landing control method of the miniature self-service giro based on behavior model, is characterized in that performing step is as follows:

(1) gather fly to control hand control the main slurry in miniature self-service giro landing stage total apart from, throttle threshold values and miniature self-service giro at the horizontal level in landing stage, highly, speed state information;

(2) build and fly to control hand behavior model, by study, fly to control hand and control behavioral data and miniature self-service giro status information, utilize adaptive fuzzy control method to build and fly to control hand behavior model, this model consists of input layer, obscuring layer, rules layer and deblurring layer, represents condition conclusion IF-THEN control procedure; Input layer be input as height error Δ height and height error variable quantity

be output as height error and height error variable quantity through normalized, input layer output is defined as:

$O_i^{\left(1\right)}=X_i,i=\mathrm{1,2}$

Wherein, X _ifor the input of input layer, be respectively height error Δ height and height error variable quantity

output for input layer; The output of input layer is as the input of obscuring layer, and the member function of the controller of obscuring layer utilizes Gaussian function definition, and obscuring layer output is defined as:

$O_{\mathrm{ik}}^{\left(2\right)}=\exp (-{(X_i-a_{\mathrm{ik}})}^2/(k_i\&CenterDot;{b^2}_{\mathrm{ik}})),i=\mathrm{1,2},k=\mathrm{1,2}...N$

Wherein, O _ikfor the output of fuzzy variable corresponding to obscuring layer, represent respectively N kind situation, x _ifor the input value after normalized, a _ikrepresent that i inputs the central value of corresponding k class Gaussian function,

represent that i inputs the mean square deviation of corresponding k class Gaussian function, k _ifor the regulation and control parameter of radial base neural net to i input, N is fuzzy variable number corresponding to input; Obscuring layer is output as height error and fuzzy variable corresponding to height error variable quantity, and obscuring layer is output as the input of rules layer, based on minimax MAX-MIN product synthetic method, generates corresponding N ²individual fuzzy rule, its output node is defined as

$O_l^{\left(3\right)}=o_{1m}^{\left(2\right)}\&cap;o_{2n}^{\left(2\right)}=\exp \left(I_{1m}^{\left(2\right)}\right)\&cap;\exp \left(I_{2n}^{\left(2\right)}\right)$

l＝N·(m-1)+nm＝1,2…N,n＝1,2…N

At deblurring layer, the numerical value that utilizes membership function weighting average decision method that rules layer is drawn is converted into the accurate output of direct control object, always apart from the controlled quentity controlled variable of rudder amount, and improve inputting and fly to control the input and output of hand behavior model by radial base neural net, improve and fly to control the adaptability of hand behavior model to environment, its output node is defined as follows

$O^{\left(4\right)}=k_3\&CenterDot;\underset{j=1}{\overset{\mathrm{ll}}{\mathrm{\&Sigma;}}}(o_j^{\left(3\right)}\&CenterDot;w_j)/\underset{j=1}{\overset{\mathrm{ll}}{\mathrm{\&Sigma;}}}{o}_j^{\left(3\right)},\mathrm{ll}=N^2$

W wherein _jthe corresponding weights of j rule, k ₃for the regulation and control parameter of radial base neural net output, O ⁽⁴⁾for always apart from the controlled quentity controlled variable of rudder amount;

(3) determine the rule of conduct in autonomous landing stage, initial period, liftoff stage and mission phase will be divided into autonomous takeoff phase, initial period is by throttle and always apart from threshold values, obtain adhesion, the liftoff stage flies to control the outer attitude of disturbing acquisition of hand behavior model elimination by self-adaptation and controls, and mission phase is realized attitude by conventional control method and controlled; The Autonomous landing stage is divided into mission phase, liftoff stage and landing stage, mission phase is realized attitude by conventional control method and is controlled, the liftoff stage flies to control by self-adaptation position and the attitude that hand behavior model and conventional control method realize and controls, and the stage that lands is realized position control by flying to control hand behavior model.

2. the autonomous landing control method of the miniature self-service giro based on behavior model according to claim 1, it is characterized in that: the described radial base neural net of step (2) carries out improved method to input and the input and output that fly to control hand behavior model and is: by sample data train RBF Neural Network, come real-time optimization to fly to control the Gaussian function mean square deviation of hand behavior model obscuring layer and always apart from the controlled quentity controlled variable of rudder amount, radial base neural net is three-layer neural network, comprise input layer, hidden layer and output layer form, horizontal level difference and the speed difference of the liftoff stage of miniature self-service giro that is input as of input layer when floating state, its output node is defined as follows

$O{O}_i^{\left(1\right)}={\mathrm{XX}}_i,i=\mathrm{1,2}$

Wherein, XX _ifor the input of input layer, be respectively horizontal level difference DELTA lateral and the speed difference DELTA speed of liftoff stage of miniature self-service giro when floating state,

output for input layer; The output that is input as input layer of hidden layer, hidden layer has Z Gaussian function to form, the input through weights stack as output layer, its output node is defined as follows

${\mathrm{OO}}_c^{\left(2\right)}=\exp ({(\underset{i=1}{\overset{2}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ic}}{\mathrm{XX}}_i-{\mathrm{aa}}_c)}^2/{\mathrm{bb}}_c^2),c=\mathrm{1,2}...Z$

Wherein, w _icbe expressed as the weights of c Gaussian function of i corresponding input, aa _cthe central value that represents c corresponding Gaussian function,

the mean square deviation that represents c corresponding Gaussian function, Z is hidden layer neuron number; Output layer is output as for adjusting the coefficient k of Gaussian function corresponding to height error ₁, for adjusting the coefficient k of Gaussian function corresponding to height error ₂, and for adjusting output always apart from the coefficient k of rudder amount ₃, its output node is defined as follows

$k_p^{\left(3\right)}=\underset{e=1}{\overset{Z}{\mathrm{\&Sigma;}}}{\mathrm{oo}}_{\mathrm{pe}}^{\left(2\right)}\&CenterDot;w_{\mathrm{pe}},p=\mathrm{1,2,3}$

Wherein, w _pefor the weights of hidden layer neuron, the output k of output layer ₁for for adjusting the coefficient of Gaussian function corresponding to height error, k ₂for adjusting the coefficient of Gaussian function corresponding to height error, k ₃for adjusting output always apart from the coefficient of rudder amount.

Images