CN106155076A - A kind of stabilized flight control method of many rotor unmanned aircrafts - Google Patents

Abstract

The invention provides the stabilized flight control method of a kind of many rotor unmanned aircrafts, comprise the steps: 1) by the flight real-time running data of airborne sensor acquisition aircraft self, and pass through carried processor the kinematics problem of aircraft is carried out corresponding dissection process, set up vehicle dynamics model；2) according to becoming ginseng convergence differential neurodynamics method for designing, the change ginseng convergence differential Neural Networks Solution device of design multi-rotor aerocraft kinetic model；3) utilize the aircraft real-time running data obtained and targeted attitude data, solve the output controlled quentity controlled variable of aircraft motor by becoming ginseng convergence differential Neural Networks Solution device；4) by step 3) solving result pass to aircraft motor governor control unmanned vehicle motion.The present invention based on becoming ginseng convergence differential neurodynamics method, can quickly, accurately and real-time approximation problem correctly solve, can the multiple time-varying problem such as solving matrix, vector, algebraically and optimization well.

Description

A kind of stabilized flight control method of many rotor unmanned aircrafts

Technical field

The invention belongs to the stabilized flight control method of many rotor unmanned aircrafts, particularly relate to a kind of based on becoming ginseng receipts Hold back the method for designing of the flight controller of differential neutral net.

Background technology

In recent years, world's unmanned vehicle technology obtains swift and violent development, has VTOL, steadily hovering, wireless biography The multi-rotor aerocraft of defeated, remote aerial photography and autonomous cruise ability has broad application prospects in military and civil field.Little Type rotary wind type unmanned plane owing to there is the mobility of excellence, simple frame for movement, convenient disposing and safeguard the features such as simple, Be widely used in taking photo by plane photography, electric inspection process, environmental monitoring, the fire prevention of deep woods, the condition of a disaster inspection, anti-probably lifesaving, military surveillance and The fields such as battle assessment.Along with the extensive application of unmanned vehicle, stablize and react the design of quick unmanned aerial vehicle (UAV) control device Cause the concern of numerous researcher, and traditional unmanned aerial vehicle (UAV) control device is all based on PID closed loop control algorithm and corresponding Improve what control algolithm was designed.Owing to PID controller and feeding back closed-loop control system design simple, and have been provided with relatively Good control effect, designs at the controller of aircraft and is widely used.Although PID controller is easy to use, but to reality The most preferably control effect then relative difficult, and be difficult to some complex task.On the basis of this, it would be desirable to design The controller that performance is better, and become ginseng convergence differential neutral net and there is super exponential convergence characteristic, it is possible to achieve aircraft Quick response, it quickly solves characteristic and is suitable for the design and development of controller of aircraft.Simultaneously as this neutral net Having stronger robustness, designed controller system is stable and controls respond well.

Summary of the invention

It is an object of the invention to overcome the deficiencies in the prior art, it is provided that a kind of based on becoming ginseng convergence differential neutral net The method for designing of stabilized flight controller.

In order to realize foregoing invention, the technical scheme of employing is as follows:

The stabilized flight control method of a kind of many rotor unmanned aircrafts, comprises the steps:

1) obtained with position sensor by attitude transducer airborne on many rotor unmanned aircrafts and corresponding height Take the flight real-time running data of aircraft self, and the processor carried by many rotor unmanned aircrafts is to aircraft Kinematics problem carries out corresponding dissection process, sets up vehicle dynamics model；

2) according to becoming ginseng convergence differential neurodynamics method for designing, the change ginseng of design multi-rotor aerocraft kinetic model Convergence differential Neural Networks Solution device；

3) utilize step 1) in the aircraft real-time running data and the targeted attitude data that obtain, by step 2) designed The ginseng convergence differential Neural Networks Solution device that becomes solve the output controlled quentity controlled variable of aircraft motor；

4) by step 3) solving result pass to aircraft motor governor control unmanned vehicle motion.

The present invention is based on becoming ginseng convergence differential neurodynamics method, and the method uses the hidden kinetic model generally existed It is described, the derivative information of each time-varying parameter can be made full use of from method and system aspect, problem solving is had necessarily Predictive ability, can quickly, accurately and real-time approximation problem correctly solve, can solving matrix, vector, algebraically and excellent well The multiple time-varying problems such as change.

Accompanying drawing explanation

Fig. 1 is the stabilized flight control method flow chart of many rotor unmanned aircrafts of the embodiment of the present invention.

Fig. 2 is the multi-rotor aerocraft structural side view of the present invention.

Fig. 3 is the multi-rotor aerocraft structure top view of the present invention.

Fig. 4 is the multi-rotor aerocraft structure three-dimensional view of the present invention.

Fig. 5 is multi-rotor aerocraft body axis system figure.

Fig. 6 is nerve network controller site layer Simulation Control effect.

Fig. 7 is nerve network controller velocity layer Simulation Control effect.

Fig. 8 is nerve network controller acceleration layer Simulation Control effect.

Detailed description of the invention

The present invention is described further below in conjunction with the accompanying drawings.

Fig. 1 is the design flow diagram of the present invention, can complete setting of aircraft nerve network controller by illustrated steps Meter:

The stabilized flight control method of a kind of many rotor unmanned aircrafts, comprises the steps:

1) obtained with position sensor by attitude transducer airborne on many rotor unmanned aircrafts and corresponding height Take the flight real-time running data of aircraft self, set up vehicle dynamics model, and by many rotor unmanned aircrafts institute The processor carried carries out corresponding dissection process to the kinematics problem of aircraft；

2) according to becoming ginseng convergence differential neurodynamics method for designing, the change ginseng of design multi-rotor aerocraft kinetic model Convergence differential Neural Networks Solution device；

3) utilize step 1) in the aircraft real-time running data and the targeted attitude data that obtain, by step 2) designed The ginseng convergence differential Neural Networks Solution device that becomes solve the output controlled quentity controlled variable of aircraft motor；

4) by step 3) solving result pass to aircraft motor governor control many rotor unmanned aircrafts motion.

Mechanism shown in Fig. 2, Fig. 3 and Fig. 4 is a kind of rotor craft structure in multi-rotor aerocraft.This structure is Six rotorcraft mechanism model, this mechanism model is by multi-rotor aerocraft propeller 1, brushless electric machine 2, rotor arm 3 and fuselage 4 Composition.The output of six of which motor is made a concerted effort and controlled quentity controlled variable u of synthesis rotating torques composition multi-rotor aerocraft₁(t)～u₄ (t).And the control design case of the present invention is to restrain differential Neural Networks Solution multi-rotor aerocraft by the designed ginseng that becomes Controlled quentity controlled variable, thus control aircraft flight, it is achieved the stability contorting of aircraft.Wherein, the rotation arrows direction in Fig. 3 and Fig. 4 The direction of rotation of indication motor, and illustrating direction of rotation clockwise is to realize mutually supporting of motor torque with combination purpose counterclockwise Disappear, it is achieved stable course changing control.

Fig. 5 show the body axis system schematic diagram at multi-rotor aerocraft place.It is following fixed to make according to body axis system Justice:

6. number the most 1. number (1), it is respectively to according to clockwise definition six motors of six rotorcraft；

(2), X-axis along 1. number rotor arm direction, point to aircraft direction of advance by body center of gravity；

(3), Y-axis along 2., the axis of symmetry direction of 3. number rotor arm, point to motion side on the right side of aircraft by body center of gravity To；

(4), Z axis is perpendicular to six rotor plane upwards, by body center of gravity sensing aircraft climb direction；

(5), pitching angle theta (t) be folded angle between body X-axis and the earth horizontal plane, setting just is downwards；

(6), roll angle φ (t) be body Z axis and cross body X-axis the earth perpendicular between angle, aircraft is to the right Shi Weizheng；

(6) during, yaw angle ψ (t) is body X-axis projection on the earth horizontal plane and earth coordinates folded between X-axis Angle, Nose Left is just.

The Simulation Control effect curve that Fig. 6, Fig. 7 and Fig. 8 are obtained by designed aircraft nerve network controller.Real Test emulation for from initial position x (0)=0, y (0)=0 and z (0)=0 and initial attitude θ (0)=0, φ (0)=0 and ψ (0)=0 starts, and utilizes the controller of aircraft based on becoming ginseng convergence differential neutral net make aircraft run and reach target position Put x (t)=0, y (t)=0 and z (t)=30 and dbjective state θ (t)=0.3, φ (t)=-0.3 and ψ (t)=0.15.God P=1 is taken through the regulatory factor p of network.Wherein Fig. 6 is site layer simulation result, and curve all can rapidly converge to dbjective state. Fig. 7 is velocity layer simulation result, this can converge to 0 again making each velocity layer parameter after control adjusts.Fig. 8 is acceleration layer Simulation result, can converge to 0 making each acceleration layer parameter after control adjusts.

According to the correlation step of design flow diagram, carry out detailed arithmetic analysis for the present invention.First, flown by above-mentioned The definition of row device attitude variable, the present invention utilizes Quaternion algebra and Kalman filtering scheduling algorithm, it is possible to achieve utilize many rotations The sensors such as gyroscope and the accelerometer carried on rotor aircraft obtain real-time attitude data θ (t) of aircraft, φ (t) with And ψ (t), utilize height sensor and position sensor to obtain aircraft position data x (t) in three dimensions, y (t) With z (t).More than complete flow chart sensing data and obtain the related content of 1.

Analyze process based on physical model above, according to different rotor craft models, set up for this aircraft Physical model equation and kinetics equation, can complete dynamic analysis by following vehicle dynamics modeling procedure:

Ignore air drag effect suffered by aircraft, physical model can be set up for aerocraft system:

$\left[\begin{array}{cc}mI& 0_{3\×3}\\ 0_{3\×3}& J\end{array}\right]\left[\begin{array}{c}\stackrel{\·}{v}\\ \stackrel{\·}{w}\end{array}\right]+\left[\begin{array}{c}w\×\left(mv\right)\\ w\×\left(Jw\right)\end{array}\right]=\left[\begin{array}{c}F\\ T\end{array}\right]+\left[\begin{array}{c}G\\ 0_{3\×1}\end{array}\right],---\left(1\right)$

Wherein m is the gross mass of aircraft, and I is 3 × 3 unit matrixs, and J is aircraft moment of inertia matrix, v and w is for flying The velocity of row device earth axes and angular velocity vector, F and G is respectively aircraft motor output axial thrust load with joint efforts and vows Amount and the axial thrust load vector of gravity, T is aircraft rotating torque vector；

Set up earth axes X_GAnd aircraft body coordinate system X_U, wherein between earth axes and body axis system There is following relation: X_U=KX_G.In transformational relation, K is the rotational transformation matrix between earth axes and body axis system, Can be expressed as

$K=\left[\begin{array}{ccc}\cos \mathrm{\θ}\left(t\right)\mathrm{\ψ}\left(t\right)& \cos \mathrm{\θ}\left(t\right)\sin \mathrm{\ψ}\left(t\right)& -\sin \mathrm{\φ}\left(t\right)\\ \sin \mathrm{\θ}\left(t\right)\cos \mathrm{\ψ}\left(t\right)\sin \mathrm{\φ}\left(t\right)-\sin \mathrm{\ψ}\left(t\right)\cos \mathrm{\φ}\left(t\right)& \sin \mathrm{\θ}\left(t\right)\sin \mathrm{\ψ}\left(t\right)\sin \mathrm{\φ}\left(t\right)+\cos \mathrm{\ψ}\left(t\right)\cos \mathrm{\φ}\left(t\right)& \cos \mathrm{\θ}\left(t\right)\sin \mathrm{\φ}\left(t\right)\\ \sin \mathrm{\θ}\left(t\right)\cos \mathrm{\ψ}\left(t\right)\cos \mathrm{\φ}\left(t\right)+\sin \mathrm{\ψ}\left(t\right)\sin \mathrm{\φ}\left(t\right)& \sin \mathrm{\θ}\left(t\right)\sin \mathrm{\ψ}\left(t\right)\cos \mathrm{\φ}\left(t\right)-\cos \mathrm{\ψ}\left(t\right)\sin \mathrm{\φ}\left(t\right)& \cos \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)\end{array}\right],$

Wherein θ (t) is the angle of pitch, and ψ (t) is yaw angle, and φ (t) is roll angle；

Theoretical according to coordinate transform, on the translation direction and rotation direction of aircraft, can according to above physical model To obtain following kinetics equation in aircraft body coordinate system

$\left\{\begin{array}{c}\stackrel{\·\·}{x}\left(t\right)=\frac{\left(\sin \mathrm{\θ}\right(t\left)\cos \mathrm{\ψ}\right(t\left)\cos \mathrm{\φ}\right(t)+\sin \mathrm{\ψ}(t\left)\sin \mathrm{\φ}\right(t\left)\right)}{m}{u}_1\left(t\right)\\ \stackrel{\·\·}{y}\left(t\right)=\frac{\left(\sin \mathrm{\θ}\right(t\left)\sin \mathrm{\ψ}\right(t\left)\cos \mathrm{\φ}\right(t)-\cos \mathrm{\φ}(t\left)\sin \mathrm{\φ}\right(t\left)\right)}{m}{u}_1\left(t\right)\\ \stackrel{\·\·}{z}\left(t\right)=\frac{\cos \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)}{m}{u}_1\left(t\right)-g\\ \stackrel{\·\·}{\mathrm{\φ}}\left(t\right)=\frac{l\·u_2\left(t\right)+(J_y-J_z)\stackrel{\·}{\mathrm{\ψ}}\left(t\right)\stackrel{\·}{\mathrm{\θ}}\left(t\right)}{J_x}\\ \stackrel{\·\·}{\mathrm{\θ}}\left(t\right)=\frac{l\·u_3\left(t\right)+(J_z-J_x)\stackrel{\·}{\mathrm{\ψ}}\left(t\right)\stackrel{\·}{\mathrm{\φ}}\left(t\right)}{J_y}\\ \stackrel{\·\·}{\mathrm{\ψ}}\left(t\right)=\frac{u_4\left(t\right)+(J_x-J_y)\stackrel{\·}{\mathrm{\θ}}\left(t\right)\stackrel{\·}{\mathrm{\φ}}\left(t\right)}{J_z}\end{array}\right.,---\left(2\right)$

Wherein x, y, z is respectively aircraft position coordinates in world coordinate system；J_x、J_yAnd J_zIt is respectively aircraft at X The rotary inertia of axle, Y-axis and Z-direction；L is brachium；G is acceleration of gravity；Synthesis controlled quentity controlled variable u₁(t)～u₄T () is by aircraft Thrust output and the synthesis torque of motor are constituted.u₁T () is making a concerted effort on aircraft vertical ascent direction, u₂T () is roll angle Make a concerted effort in direction, u₃T () is to make a concerted effort in angle of pitch direction, u₄T () is yaw angle direction composition torque.

Specifically, described step 2) specifically include:

By becoming ginseng convergence differential neurodynamics method for designing, respectively by Z axis height z (t), roll angle φ (t), pitching Angle θ (t) and yaw angle ψ (t) set out, and design about output controlled quentity controlled variable u₁(t)～u₄T the ginseng that becomes of () restrains differential neutral net Systematic parameter departure function；

Respectively according to calculated about output controlled quentity controlled variable u₁(t)～u₄The system becoming ginseng convergence differential neutral net of (t) Parameter error function, design becomes ginseng convergence differential Neural Networks Solution device.

Wherein, described by becoming ginseng convergence differential neurodynamics method for designing, respectively by Z axis height z (t), roll angle φ (t), pitching angle theta (t) and yaw angle ψ (t) set out, and design about output controlled quentity controlled variable u₁(t)～u₄The change ginseng convergence differential of (t) The step of the systematic parameter departure function of neutral net specifically includes；

For Z axis height z (t), according to the target setting height value z in Z-direction_TAnd actual height value z (t), Can define on site layer about actual height value z (t) departure function e_z1T () is: e_z1(t)=z (t)-z_T.In order to make actual value Z (t) can converge to desired value z_T, according to becoming ginseng convergence differential neurodynamics method for designing, can design based on deviation letter The neurodynamics equation of numberWherein γ (t)=t^p+ p is time-varying parameter, represents convergency factor Regulatory factor.According to departure function e_z1(t)=z (t)-z_T, can obtainBy e_z1(t)=z (t)-z_TWith AndSubstitute intoCan obtainNamely

$\stackrel{\·}{z}\left(t\right)+(t^p+p)\left(z\right(t)-z_T)=0;---\left(3\right)$

SolveUnderstand According to departure function e_z1(t)=z (t)-z_T, it is known thatOn if State formula to set up, then, when reaching dbjective state, site layer z (t) will restrain target z_T；According to practical situation, at aircraft After reaching specified altitude assignment, aircraft is in the speed of vertical directionIt is necessarily equal to 0；Simultaneously as equation (3) do not comprise about Controlled quentity controlled variable u₁(t)～u₄The relevant information of (t), it is impossible to realize controlled quentity controlled variable is solved, therefore need further design packet to contain speed LayerAnd acceleration layerDeparture function, then define According to becoming ginseng convergence differential neutral net method for designing, kinetics equation based on departure function can be designedWherein γ (t)=t^p+ p, according to Known deviation function e_z2The derivative of (t)For: Will be above with respect to e_z2(t) andEquation substitute into equationCan obtain such as minor function Formula: Namely

$\stackrel{\·\·}{z}\left(t\right)+2(t^p+p)\stackrel{\·}{z}\left(t\right)+(t^{2p}+2{\mathrm{pt}}^p+{\mathrm{pt}}^{p-1}+p^2)\left(z\right(t)-z_T)=0;---\left(4\right)$

Solve the differential equationHave According toWith Can obtain Understand0 will be converged to；

When equation (4) is set up, velocity layerTo converge to 0 and site layer z (t) target z will be restrained_T.Accordingly, may be used To consider departure function

$E_Z\left(t\right)=\stackrel{\·\·}{z}\left(t\right)+2(t^p+p)\stackrel{\·}{z}\left(t\right)+(t^{2p}+2{\mathrm{pt}}^p+{\mathrm{pt}}^{p-1}+p^2)\left(z\right(t)-z_T),---\left(5\right)$

For obtaining the realistic model of neutral net, binding kinetics equation (2), equation (5) can be written over Further, in order to simplify formula, departure function E_zCan be rewritten as

E_z(t)=a_z(t)u₁(t)-b_z(t), (6)

Wherein Namely obtain about output controlled quentity controlled variable u₁The departure function of (t)；

For roll angle φ (t), for the angle on target φ reached_T, first definition error function e_φ1(t)=φ (t)- φ_TAndOwing to solving at angle layer, can obtain according to becoming ginseng differential neurodynamics method for designingWherein γ (t)=t^p+ p is time-varying parameter, represents convergency factor regulatory factor；By error function e_φ1(t)=φ (t)-φ_TWithSubstitute into equation Can obtain

$\stackrel{\·}{\mathrm{\φ}}\left(t\right)+(t^p+p)\left(\mathrm{\φ}\right(t)-{\mathrm{\φ}}_T)=0,---\left(7\right)$

According to the differential equationCan solve According to e_φ1(t)=φ (t)-φ_T, can obtainUnderstanding φ (t) will be with super exponential convergence to mesh Mark angle φ_T；Due to knownNeeds solve containingEquation；In order to be contained's Equation, utilizes same method to set error functionAnd Further according to becoming ginseng convergence differential neurodynamics design side MethodWherein γ (t)=t^p+p；Solve the differential equation Can To obtain deviation

Understand simultaneouslyAvailableSolution beAnd

Can obtain

According to solving knot Really, it is known that final0 will be converged to.By error functionWithSubstitute into neurodynamics design formulaCan obtain

Namely

This constraint is equivalent to lower deviation such as and converges to zero, i.e.

When departure function converges to 0, aircraft reaches dbjective state.According to described kinetic model equation, deviation Function can be further converted to

$E_{\mathrm{\φ}}\left(t\right)=\frac{l}{J_x}\·u_2\left(t\right)+b_{\mathrm{\φ}}\left(t\right),---\left(10\right)$

Wherein Namely obtain about output controlled quentity controlled variable u₃The departure function of (t)；

For pitching angle theta (t), in order to reach angle on target θ_T, first definition error function e_θ1(t)=θ (t)-θ_TAndOwing to solving at angle layer, can obtain according to becoming ginseng differential neurodynamics method for designingWherein γ (t)=t^p+ p is time-varying parameter, represents convergency factor regulatory factor；By error functionWithSubstitute into equationCan obtain

$\stackrel{\·}{\mathrm{\θ}}\left(t\right)+(t^p+p)\left(\mathrm{\θ}\right(t)-{\mathrm{\θ}}_T)=0,---\left(11\right)$

According to the differential equationCan solve According to e_θ1(t)=θ (t)-θ_T, can obtainUnderstanding θ (t) will be with super exponential convergence To angle on target θ_T.Due to knownNeeds solve containingEquation；In order to be contained's Equation, utilizes same method to set error functionAnd Further according to becoming ginseng convergence differential neurodynamics method for designingWherein γ (t)=t^p+p；Solve the differential equationPermissible Obtain deviation Understand simultaneouslyAvailableSolution beAndCan obtain According to solving result, it is known that finalTo converge to 0；By error functionWith Substitute into neurodynamics design formulaCan obtainThe most just It is

This constraint is equivalent to lower deviation such as and converges to zero, i.e.

When departure function converges to 0, aircraft reaches dbjective state.According to the kinetic model side in claim 2 Journey, departure function can be further converted to

$E_{\mathrm{\θ}}\left(t\right)=\frac{l}{J_y}\·u_3\left(t\right)+b_{\mathrm{\θ}}\left(t\right),---\left(14\right)$

Wherein Namely obtain about output controlled quentity controlled variable u₃The departure function of (t)；

For yaw angle ψ (t), for the angle on target ψ reached_T, first definition error function e_ψ1(t)=ψ (t)-ψ_TAndOwing to solving at angle layer, can obtain according to becoming ginseng differential neurodynamics method for designingWherein γ (t)=t^p+ p is time-varying parameter, represents convergency factor regulatory factor；By error function e_ψ1(t)=ψ (t)-ψ_TWithSubstitute into equation Can obtain

$\stackrel{\·}{\mathrm{\ψ}}\left(t\right)+(t^p+p)\left(\mathrm{\ψ}\right(t)-{\mathrm{\ψ}}_T)=0,---\left(15\right)$

According to the differential equationCan solve According to e_ψ1(t)=ψ (t)-ψ_T, can obtainUnderstanding ψ (t) will be with super exponential convergence to mesh Mark angle ψ_T；Due to knownNeeds solve containingEquation；In order to be contained's Equation, utilizes same method to set error functionAnd Further according to becoming ginseng convergence differential neurodynamics design side MethodWherein γ (t)=t^p+p.Solve the differential equation Can To obtain deviation

Understand simultaneouslyAvailableSolution ForAndCan obtain

According to solving result, it is known that final0 will be converged to；By error function WithSubstitute into nerve dynamic Mechanics design formulaCan obtain Namely

This constraint is equivalent to lower deviation such as and converges to zero, i.e.

When departure function converges to 0, aircraft reaches dbjective state.According to the kinetic model side in claim 2 Journey, departure function can be further converted to

$E_{\mathrm{\ψ}}\left(t\right)=\frac{1}{J_z}\·u_4\left(t\right)+b_{\mathrm{\ψ}}\left(t\right),---\left(18\right)$

Wherein Namely obtain about output controlled quentity controlled variable u₄The departure function of (t).

Described respectively according to calculated about output controlled quentity controlled variable u₁(t)～u₄T the ginseng that becomes of () restrains differential neutral net Systematic parameter departure function, design becomes the step of ginseng convergence differential Neural Networks Solution device and specifically includes:

For Z axis height z (t), utilize and become ginseng convergence differential neutral net method for designing, can designBy equation (6) and its derivative Substitute into equationHave

According to above-mentioned equation with And a_Z(t), b_zT the expression formula of (), can write out complete neural network structure expression formula

$\begin{array}{c}{\stackrel{\·}{u}}_1\left(t\right)=\\ \left(\right((t^p+p)\frac{\cos \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)}{m}+\frac{\cos \mathrm{\θ}\left(t\right)\sin \mathrm{\φ}\left(t\right)}{m}\stackrel{\·}{\mathrm{\φ}}\left(t\right)+\frac{\sin \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)}{m}\stackrel{\·}{\mathrm{\θ}}\left(t\right)\left)u_1\right(t)-\\ (t^p+p)g+(-t^{2p}-2{\mathrm{pt}}^p+3{\mathrm{pt}}^{p-1}-p^2)\stackrel{\·}{z}\left(t\right)+(-t^{3p}-3{\mathrm{pt}}^{2p}+{\mathrm{pt}}^{2p-1}-\\ 3p^2{t}^p+p^2{t}^{p-1}+p(p-1)t^{p-2}-p^3)\left(z\right(t)-z_T))\·\frac{m}{\cos \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)}\end{array},---\left(19\right)$

To sum up can obtain becoming the implicit expression kinetics equation of ginseng convergence differential neutral net；According toHaveSubstitute into equation (6) to have

$u_1\left(t\right)=\left(\exp \right(\frac{1}{p+1}{t}^{p+1}+pt\left)\right(a_Z\left(0\right)u_1\left(0\right)-b_Z\left(0\right))+b_Z(t\left)\right)\·\frac{1}{a_Z\left(t\right)},---\left(20\right)$

According to the characteristic of this neutral net, position z (t) will converge to target location z with hyperexponential form_TWithWill Converge to 0；

For roll angle φ (t), according to becoming ginseng convergence differential neurodynamics method for designing, can designNamelyAccording to Above-mentioned equation, has

Solve the differential equationHave WhereinTherefore have

$u_2\left(t\right)=\left(\exp \right(\frac{1}{p+1}{t}^{p+1}+pt\left)\right(\frac{l}{J_x}\·u_2\left(0\right)+b_{\mathrm{\φ}}\left(0\right))-b_{\mathrm{\φ}}(t\left)\right)\·\frac{J_x}{l},---\left(22\right)$

Also the solution of equation (21) it is；

For pitching angle theta (t), according to becoming ginseng convergence differential neurodynamics method for designing, can designNamelyAccording to upper State equation, have

Solve the differential equationHave WhereinTherefore have

$u_3\left(t\right)=\left(\exp \right(\frac{1}{p+1}{t}^{p+1}+pt\left)\right(\frac{l}{J_y}\·u_3\left(0\right)+b_{\mathrm{\θ}}\left(0\right))-b_0(t\left)\right)\·\frac{J_y}{l},---\left(24\right)$

Also the solution of equation (23) it is；

For yaw angle ψ (t), according to becoming ginseng convergence differential neurodynamics method for designing, can designNamelyAccording to Above-mentioned equation, has

Solve the differential equationHave WhereinTherefore have

$u_4\left(t\right)=\left(\exp \right(\frac{1}{p+1}{t}^{p+1}+pt\left)\right(\frac{1}{J_z}\·u_4\left(0\right)+b_{\mathrm{\ψ}}\left(0\right))-b_{\mathrm{\ψ}}(t\left)\right)\·J_z,---\left(26\right)$

Also the solution of equation (25) it is；

Utilize neutral net equation (20), (22), (24) and (26), solve synthesis controlled quentity controlled variable u₁(t)～u₄T () is The controlled quentity controlled variable of corresponding aircraft flight demand, in like manner, can obtain controlled quentity controlled variable u according to z (t)₁The neutral net equation of (t), root U can be obtained according to θ (t)₃T (), can obtain u according to ψ (t)₄(t), concrete as follows:

Wherein

Controlled quentity controlled variable u that will solve₁(t)～u₄T () is carried out not according to structure and the motor number of different rotor crafts Same output controls distribution.

According to controlled quentity controlled variable u calculated by above-mentioned Neural Networks Solution process₁(t)～u₄T (), for the knot of different aircraft Structure and motor number, realized the control of each motor, to sum up complete flow chart motor by corresponding motor control allocation Control allocation and motor control.The present invention can be completed according to above-mentioned steps.

The above embodiment of the present invention is only for clearly demonstrating example of the present invention, and is not to the present invention The restriction of embodiment.For those of ordinary skill in the field, can also make on the basis of the above description The change of other multi-form or variation.Here without also cannot all of embodiment be given exhaustive.All the present invention's Any amendment, equivalent and the improvement etc. made within spirit and principle, should be included in the protection of the claims in the present invention Within the scope of.

Claims (5)

1. the stabilized flight control method of rotor unmanned aircraft more than a kind, it is characterised in that comprise the steps:

1) flown with position sensor acquisition by airborne attitude transducer on many rotor unmanned aircrafts and corresponding height The flight real-time running data of row device self, is set up vehicle dynamics model, and is carried by many rotor unmanned aircrafts Processor the kinematics problem of aircraft is carried out corresponding dissection process；

2) according to becoming ginseng convergence differential neurodynamics method for designing, the change ginseng convergence of design multi-rotor aerocraft kinetic model Differential Neural Networks Solution device；

3) utilize step 1) in the aircraft real-time running data and the targeted attitude data that obtain, by step 2) designed by change Ginseng convergence differential Neural Networks Solution device solves the output controlled quentity controlled variable of aircraft motor；

4) by step 3) solving result pass to aircraft motor governor control many rotor unmanned aircrafts motion.

The stabilized flight control method of a kind of many rotor unmanned aircrafts the most according to claim 1, it is characterised in that institute State and set up vehicle dynamics model, and the motion knowledge that the processor carried by many rotor unmanned aircrafts is to aircraft Topic carries out the step of corresponding dissection process and specifically includes:

Ignore air drag effect suffered by aircraft, physical model can be set up for aerocraft system:

$\left[\begin{array}{cc}mI& 0_{3\×3}\\ 0_{3\×3}& J\end{array}\right]\left[\begin{array}{c}\stackrel{\·}{v}\\ \stackrel{\·}{w}\end{array}\right]+\left[\begin{array}{c}w\×\left(mv\right)\\ w\×\left(Jw\right)\end{array}\right]=\left[\begin{array}{c}F\\ T\end{array}\right]+\left[\begin{array}{c}G\\ 0_{3\×1}\end{array}\right],---\left(1\right)$

Wherein m is the gross mass of aircraft, and I is 3 × 3 unit matrixs, and J is aircraft moment of inertia matrix, v and w is aircraft The velocity of earth axes and angular velocity vector, F and G is respectively aircraft motor output axial thrust load vector with joint efforts The axial thrust load vector of gravity, T is aircraft rotating torque vector；

Set up earth axes X_GAnd aircraft body coordinate system X_U, wherein exist between earth axes and body axis system Following relation: X_U=KX_G, in transformational relation, K is the rotational transformation matrix between earth axes and body axis system, permissible It is expressed as

$K=\left[\begin{array}{ccc}\cos \mathrm{\θ}\left(t\right)\cos \mathrm{\ψ}\left(t\right)& \cos \mathrm{\θ}\left(t\right)\sin \mathrm{\ψ}\left(t\right)& -\sin \mathrm{\θ}\left(t\right)\\ \sin \mathrm{\θ}\left(t\right)\cos \mathrm{\ψ}\left(t\right)\sin \mathrm{\φ}\left(t\right)-\sin \mathrm{\ψ}\left(t\right)\cos \mathrm{\φ}\left(t\right)& \sin \mathrm{\θ}\left(t\right)\sin \mathrm{\ψ}\left(t\right)\sin \mathrm{\φ}\left(t\right)+\cos \mathrm{\ψ}\left(t\right)\cos \mathrm{\φ}\left(t\right)& \cos \mathrm{\θ}\left(t\right)\sin \mathrm{\φ}\left(t\right)\\ \sin \mathrm{\θ}\left(t\right)\cos \mathrm{\ψ}\left(t\right)\cos \mathrm{\φ}\left(t\right)+\sin \mathrm{\ψ}\left(t\right)\sin \mathrm{\φ}\left(t\right)& \sin \mathrm{\θ}\left(t\right)\sin \mathrm{\ψ}\left(t\right)\cos \mathrm{\φ}\left(t\right)-\cos \mathrm{\ψ}\left(t\right)\sin \mathrm{\φ}\left(t\right)& \cos \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)\end{array}\right],$

Wherein θ (t) is the angle of pitch, and ψ (t) is yaw angle, and φ (t) is roll angle；

Theoretical according to coordinate transform, on the translation direction and rotation direction of aircraft, can obtain according to above physical model Obtain the kinetics equation in aircraft body coordinate system as follows

$\left\{\begin{array}{c}\stackrel{\·\·}{x}\left(t\right)=\frac{\left(\sin \mathrm{\θ}\right(t\left)\cos \mathrm{\ψ}\right(t\left)\cos \mathrm{\φ}\right(t)+\sin \mathrm{\ψ}(t\left)\sin \mathrm{\φ}\right(t\left)\right)}{m}{u}_1\left(t\right)\\ \stackrel{\·\·}{y}\left(t\right)=\frac{\left(\sin \mathrm{\θ}\right(t\left)\sin \mathrm{\ψ}\right(t\left)\cos \mathrm{\φ}\right(t)-\cos \mathrm{\ψ}(t\left)\sin \mathrm{\φ}\right(t\left)\right)}{m}{u}_1\left(t\right)\\ \stackrel{\·\·}{z}\left(t\right)=\frac{\cos \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)}{m}{u}_1\left(t\right)-g\\ \stackrel{\·\·}{\mathrm{\φ}}\left(t\right)=\frac{l\·u_2\left(t\right)+(J_y-J_z)\stackrel{\·}{\mathrm{\ψ}}\left(t\right)\stackrel{\·}{\mathrm{\θ}}\left(t\right)}{J_x}\\ \stackrel{\·\·}{\mathrm{\θ}}\left(t\right)=\frac{l\·u_3\left(t\right)+(J_z-J_x)\stackrel{\·}{\mathrm{\ψ}}\left(t\right)\stackrel{\·}{\mathrm{\φ}}\left(t\right)}{J_y}\\ \stackrel{\·}{\mathrm{\ψ}}\left(t\right)=\frac{u_4\left(t\right)+(J_x-J_y)\stackrel{\·}{\mathrm{\θ}}\left(t\right)\stackrel{\·}{\mathrm{\φ}}\left(t\right)}{J_z}\end{array}\right.,---\left(2\right)$

Wherein x, y, z is respectively aircraft position coordinates in world coordinate system；J_x、J_yAnd J_zIt is respectively aircraft at X-axis, Y Axle and the rotary inertia of Z-direction；L is brachium；G is acceleration of gravity；Synthesis controlled quentity controlled variable u₁(t)～u₄T () is by aircraft motor Thrust output and synthesis torque constitute, u₁T () is making a concerted effort on aircraft vertical ascent direction, u₂T () is roll angle direction Make a concerted effort, u₃T () is to make a concerted effort in angle of pitch direction, u₄T () is yaw angle direction composition torque.

The stabilized flight control method of a kind of many rotor unmanned aircrafts the most according to claim 2, it is characterised in that institute State step 2) specifically include:

By becoming ginseng convergence differential neurodynamics method for designing, respectively by Z axis height z (t), roll angle φ (t), pitching angle theta T () and yaw angle ψ (t) set out, design about output controlled quentity controlled variable u₁(t)～u₄The system becoming ginseng convergence differential neutral net of (t) Parameter error function；

Respectively according to calculated about output controlled quentity controlled variable u₁(t)～u₄The systematic parameter becoming ginseng convergence differential neutral net of (t) Departure function, design becomes ginseng convergence differential Neural Networks Solution device.

The stabilized flight control method of a kind of many rotor unmanned aircrafts the most according to claim 3, it is characterised in that: institute State by becoming ginseng convergence differential neurodynamics method for designing, respectively by Z axis height z (t), roll angle φ (t), pitching angle theta (t) Set out with yaw angle ψ (t), design about output controlled quentity controlled variable u₁(t)～u₄The system ginseng becoming ginseng convergence differential neutral net of (t) The step of number departure function specifically includes:

For Z axis height z (t), according to the target setting height value z in Z-direction_TAnd actual height value z (t), at site layer On can define about actual height value z (t) departure function e_z1T () is: e_z1(t)=z (t)-z_T, in order to make actual value z (t) energy Enough converge to desired value z_T, according to becoming ginseng convergence differential neurodynamics method for designing, god based on departure function can be designed Through kinetics equationWherein γ (t)=t^p+ p is time-varying parameter, represent convergency factor regulation because of Son；According to departure function e_z1(t)=z (t)-z_T, can obtainBy e_z1(t)=z (t)-z_TAndSubstitute intoCan obtainNamely

$\stackrel{\·}{z}\left(t\right)+(t^p+p)\left(z\right(t)-z_T)=0;---\left(3\right)$

SolveUnderstand According to departure function e_z1(t)=z (t)-z_T, it is known thatOn if State formula to set up, then, when reaching dbjective state, site layer z (t) will restrain target z_T；According to practical situation, at aircraft After reaching specified altitude assignment, aircraft is in the speed of vertical directionIt is necessarily equal to 0；Simultaneously as equation (3) do not comprise about Controlled quentity controlled variable u₁(t)～u₄The relevant information of (t), it is impossible to realize controlled quentity controlled variable is solved, therefore need further design packet to contain speed LayerAnd acceleration layerDeparture function, then define According to becoming ginseng convergence differential neutral net method for designing, kinetics equation based on departure function can be designedWherein γ (t)=t^p+ p, according to Known deviation function e_z2The derivative of (t)For: Will be above with respect to e_z2(t) andEquation substitute into equationCan obtain such as minor function Formula: Namely

$\stackrel{\·\·}{z}\left(t\right)+2(t^p+p)\stackrel{\·}{z}\left(t\right)+(t^{2p}+2{\mathrm{pt}}^p+{\mathrm{pt}}^{p-1}+p^2)\left(z\right(t)-z_T)=0;---\left(4\right)$

Solve the differential equationHave According toWith Can obtain Understand0 will be converged to；

When equation (4) is set up, velocity layerTo converge to 0 and site layer z (t) target z will be restrained_T, accordingly, Ke Yikao Consider departure function

$E_Z\left(t\right)=\stackrel{\·\·}{z}\left(t\right)+2(t^p+p)\stackrel{\·}{z}\left(t\right)+(t^{2p}+2{\mathrm{pt}}^p+{\mathrm{pt}}^{p-1}+p^2)\left(z\right(t)-z_T),---\left(5\right)$

For obtaining the realistic model of neutral net, binding kinetics equation (2), equation (5) can be written over Further, in order to simplify formula, departure function E_ZCan be rewritten as

E_Z(t)=a_Z(t)u₁(t)-b_Z(t), (6)

Wherein Namely obtain about output controlled quentity controlled variable u₁The departure function of (t)；

For roll angle φ (t), for the angle on target φ reached_T, first definition error function e_φ1(t)=φ (t)-φ_TAndOwing to solving at angle layer, can obtain according to becoming ginseng differential neurodynamics method for designingWherein γ (t)=t^p+ p is time-varying parameter, represents convergency factor regulatory factor；By error function e_φ1(t)=φ (t)-φ_TWithSubstitute into equation Can obtain

$\stackrel{\·}{\mathrm{\φ}}\left(t\right)+(t^p+P)\left(\mathrm{\φ}\right(t)-{\mathrm{\φ}}_T)=0,---\left(7\right)$

According to the differential equationCan solve According to e_φ1(t)=φ (t)-φ_T, can obtainUnderstanding φ (t) will be with super exponential convergence to target angle Degree φ_T；Due to knownNeeds solve containingEquation；In order to be containedEquation, utilize same Method sets error functionAnd Further according to becoming ginseng convergence differential neurodynamics design side MethodWherein γ (t)=t^p+p；Solve the differential equation Can To obtain deviation

Understand simultaneouslyAvailableSolution beAnd

Can obtain

According to asking Solve result, it is known that final0 will be converged to；By error functionWithSubstitute into neurodynamics design formulaCan obtain

Namely

This constraint is equivalent to lower deviation such as and converges to zero, i.e.

When departure function converges to 0, aircraft reaches dbjective state, according to described kinetic model equation, departure function Can be further converted to

$E_{\mathrm{\φ}}\left(t\right)=\frac{l}{J_x}\·u_2\left(t\right)+b_{\mathrm{\φ}}\left(t\right),---\left(10\right)$

Wherein

Namely obtain about output controlled quentity controlled variable u₂The departure function of (t)；

For pitching angle theta (t), in order to reach angle on target θ_T, first definition error function e_θ1(t)=θ (t)-θ_TAndOwing to solving at angle layer, can obtain according to becoming ginseng differential neurodynamics method for designingWherein γ (t)=t^p+ p is time-varying parameter, represents convergency factor regulatory factor；By error function e_θ1(t)=θ (t)-θ_TWithSubstitute into equationCan obtain

$\stackrel{\·}{\mathrm{\θ}}\left(t\right)+(t^p+p)\left(\mathrm{\θ}\right(t)-{\mathrm{\θ}}_T)=0,---\left(11\right)$

According to the differential equationCan solve According to e_θ1(t)=θ (t)-θ_T, can obtainUnderstanding θ (t) will be with super exponential convergence to mesh Mark angle, θ_T；Due to knownNeeds solve containingEquation；In order to be containedSide Journey, utilizes same method to set error functionAnd Further according to becoming ginseng convergence differential neurodynamics method for designingWherein γ (t)=t^p+p；Solve the differential equationPermissible Obtain deviation Understand simultaneouslyAvailableSolution beAndCan obtain According to solving result, it is known that finalTo converge to 0；By error functionWith Substitute into neurodynamics design formulaCan obtainNamely

This constraint is equivalent to lower deviation such as and converges to zero, i.e.

When departure function converges to 0, aircraft reaches dbjective state；According to the kinetic model equation in claim 2, partially Difference function can be further converted to

$E_{\mathrm{\θ}}\left(t\right)=\frac{l}{J_y}\·u_3\left(t\right)+b_{\mathrm{\θ}}\left(t\right),---\left(14\right)$

Wherein,

Namely obtain about output controlled quentity controlled variable u₃The departure function of (t)；

For yaw angle ψ (t), for the angle on target ψ reached_T, first definition error function e_ψ1(t)=ψ (t)-ψ_TAndOwing to solving at angle layer, can obtain according to becoming ginseng differential neurodynamics method for designingWherein γ (t)=t^p+ p is time-varying parameter, represents convergency factor regulatory factor；By error function e_ψ1(t)=ψ (t)-ψ_TWithSubstitute into equation Can obtain

$\stackrel{\·}{\mathrm{\ψ}}\left(t\right)+(t^p+p)\left(\mathrm{\ψ}\right(t)-{\mathrm{\ψ}}_T)=0,---\left(15\right)$

According to the differential equationCan solve According to e_ψ1(t)=ψ (t)-ψ_T, can obtainUnderstanding ψ (t) will be with super exponential convergence to mesh Mark angle ψ_T；Due to knownNeeds solve containingEquation；In order to be containedSide Journey, utilizes same method to set error functionAnd Further according to becoming ginseng convergence differential neurodynamics design side MethodWherein γ (t)=t^p+p；Solve the differential equation Can To obtain deviation

Understand simultaneouslyAvailable's Xie Wei AndCan obtain

According to solving result, it is known that final0 will be converged to；By error function WithSubstitute into nerve dynamic Mechanics design formulaCan obtain Namely

This constraint is equivalent to lower deviation such as and converges to zero, i.e.

When departure function converges to 0, aircraft reaches dbjective state；According to the kinetic model equation in claim 2, partially Difference function can be further converted to

$E_{\mathrm{\ψ}}\left(t\right)=\frac{1}{J_z}\·u_4\left(t\right)+b_{\mathrm{\ψ}}\left(t\right),---\left(18\right)$

Wherein,

Namely obtain about output controlled quentity controlled variable u₄The departure function of (t).

The stabilized flight control method of many rotor unmanned aircrafts the most according to claim 4, it is characterised in that described point Not according to calculated about output controlled quentity controlled variable u₁(t)～u₄The systematic parameter deviation letter becoming ginseng convergence differential neutral net of (t) Number, design becomes the step of ginseng convergence differential Neural Networks Solution device and specifically includes:

For Z axis height z (t), utilize and become ginseng convergence differential neutral net method for designing, can designBy equation (6) and its derivative Substitute into equationHave

According to above-mentioned equation and a_Z (t), b_ZT the expression formula of (), can write out complete neural network structure expression formula

$\begin{array}{c}{\stackrel{\·}{u}}_1\left(t\right)=\\ \left(\right((t^p+p)\frac{\cos \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)}{m}+\frac{\cos \mathrm{\θ}\left(t\right)\sin \mathrm{\φ}\left(t\right)}{m}\stackrel{\·}{\mathrm{\φ}}\left(t\right)+\frac{\sin \mathrm{\θ}\left(t\right)\cos \mathrm{\φ}\left(t\right)}{m}\stackrel{\·}{\mathrm{\θ}}\left(t\right)\left)u_1\right(t)-\end{array}$

$\begin{array}{c}(t^p+p)g+(-t^{2p}-2{\mathrm{pt}}^p+3{\mathrm{pt}}^{p-1}-p^2)\stackrel{\·}{z}\left(t\right)+(-t^{3p}-3{\mathrm{pt}}^{2p}+{\mathrm{pt}}^{2p-1}-\\ 3p^2{t}^p+p^2{t}^{p-1}+p(p-1)t^{p-2}-p^3)\left(z\right(t)-z_T))\·\frac{m}{cos\mathrm{\θ}(t-)cos\mathrm{\φ}\left(t\right)}\end{array},---\left(19\right)$

To sum up can obtain becoming the implicit expression kinetics equation of ginseng convergence differential neutral net；According to HaveSubstitute into equation (6) to have

$u_1\left(t\right)=\left(\exp \right(\frac{1}{p+1}{t}^{p+1}+pt\left)\right({\mathrm{\α}}_Z\left(0\right)u_1\left(0\right)-b_Z\left(0\right))+b_Z\left(t\right))\·\frac{1}{a_Z\left(t\right)},---\left(20\right)$

According to the characteristic of this neutral net, position z (t) will converge to target location z with hyperexponential form_TWithWill convergence To 0；

For roll angle φ (t), according to becoming ginseng convergence differential neurodynamics method for designing, can designNamelyAccording to Above-mentioned equation, has

Solve the differential equationHave WhereinTherefore have

$u_2\left(t\right)=\left(\exp \right(\frac{1}{p+1}{t}^{p+1}+pt\left)\right(\frac{1}{J_x}\·u_2\left(0\right)+b_{\mathrm{\φ}}\left(0\right))-b_{\mathrm{\φ}}(t\left)\right)\·\frac{J_x}{l},---\left(22\right)$

Also the solution of equation (21) it is；

For pitching angle theta (t), according to becoming ginseng convergence differential neurodynamics method for designing, can designNamelyAccording to upper State equation, have

Solve the differential equationHave WhereinTherefore have

$u_3\left(t\right)=\left(\exp \right(\frac{1}{p+1}{t}^{p+1}+pt\left)\right(\frac{1}{J_y}\·u_3\left(0\right)+b_{\mathrm{\θ}}\left(0\right))-b_{\mathrm{\θ}}(t\left)\right)\·\frac{J_y}{l},---\left(24\right)$

Also the solution of equation (23) it is；

For yaw angle ψ (t), according to becoming ginseng convergence differential neurodynamics method for designing, can designNamelyAccording to Above-mentioned equation, has

Solve the differential equationHave WhereinTherefore have

$u_4\left(t\right)=\left(\exp \right(\frac{1}{p+1}{t}^{p+1}+pt\left)\right(\frac{1}{J_z}\·u_4\left(0\right)+b_{\mathrm{\ψ}}\left(0\right))-b_{\mathrm{\ψ}}\left(t\right))\·J_z,---\left(26\right)$

Also the solution of equation (25) it is；

Utilize neutral net equation (20), (22), (24) and (26), solve synthesis controlled quentity controlled variable u₁(t)～u₄T () is correspondence The controlled quentity controlled variable of aircraft flight demand, in like manner, can obtain controlled quentity controlled variable u according to z (t)₁T the neutral net equation of (), according to θ T () can obtain u₃T (), can obtain u according to ψ (t)₄(t), concrete as follows:

Wherein

Controlled quentity controlled variable u that will solve₁(t)～u₄T () carries out different according to structure and the motor number of different rotor crafts Output controls distribution.