Hindawi Publishing CorporationJournal of Applied MathematicsVolume 2013, Article ID 452653, 12 pageshttp://dx.doi.org/10.1155/2013/452653Research ArticleAdaptive Neural Network Dynamic Inversion with PrescribedPerformance for Aircraft Flight ControlWendong Gai,1,2 Honglun Wang,2 Jing Zhang,1 and Yuxia Li112College of Information and Electrical Engineering, Shandong University of Science and Technology, Qingdao 266590, ChinaScience and Technology on Aircraft Control Laboratory, Beihang University, Beijing 100191, ChinaCorrespondence should be addressed to Honglun Wang; hl wang 2002@126.comReceived 16 July 2013; Revised 23 September 2013; Accepted 4 October 2013Academic Editor: Dewei LiCopyright 2013 Wendong Gai et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.An adaptive neural network dynamic inversion with prescribed performance method is proposed for aircraft flight control. Theaircraft nonlinear attitude angle model is analyzed. And we propose a new attitude angle controller design method based onprescribed performance which describes the convergence rate and overshoot of the tracking error. Then the model error iscompensated by the adaptive neural network. Subsequently, the system stability is analyzed in detail. Finally, the proposed methodis applied to the aircraft attitude tracking control system. The nonlinear simulation demonstrates that this method can guaranteethe stability and tracking performance in the transient and steady behavior.1. IntroductionFlight control design for aircraft continues to be one ofthe most important problems in the world of automaticcontrol. The problem is driven by the nonlinear and uncertainnature of aircraft dynamics. Traditionally, the solution to thisproblem is to design the linear controller using linearizedaircraft models at multiple trimmed conditions. And thisprocedure is time consuming and expensive.Control of aircraft by dynamic model inversion is wellknown and has been applied to the control of high angleof attack fighter aircraft [1, 2]. The primary drawback ofdynamic inversion for aircraft flight control is the needfor high-fidelity nonlinear model which must be invertedin real time. However, it is difficult to obtain the exactaircraft dynamic model in practice. The neural networkaugmented model inversion in the attitude angular loop isimplemented to compensate the model inversion error, and ituses proportional-derivative desired dynamics to design theattitude control system for the helicopter [3] and tilt-rotoraircraft [4].The asymptotic tracking can be achieved using thismethod. However, the transient behavior of the output signalscould be oscillatory when the tracking error magnitude isdecreased by increasing the adaption rate. Several solutions[5โ8] have been proposed to overcome this problem. Thesemethods guarantee the convergence of tracking error, butthe required tracking error upper bounds canโt be accuratelycomputed. A new adaptive control method with prescribedperformance is presented in [9], and this method guaranteesthe transient state tracking error in the prespecified performance bound. And this method is used to improve theperformance of the planar two-link articulated manipulator[10, 11] and the 6-DOF PUMA 560 arm [12].It is very important for aircraft to track the attitude command with a desired transient and steady performance, whenthe aircraft finishes the special flight tasks, such as automatedaerial refueling [13, 14] and transition flight control [15, 16].In this paper, we will investigate the aircraft attitudecontrol problem of guaranteeing transient and steady performance in the adaptive compensation control system. Byemploying the prescribed performance bounds proposed in[9], we propose a new adaptive neural network dynamicinversion method. With certain transformation method, anew transformed error system is obtained through considering the prescribed performance bound into the originalattitude control system. An adaptive dynamic inversion controller is designed for the transformed system. It is ensured
2Journal of Applied Mathematicsthat the tracking error is guaranteed inside the prescribederror bound as long as the transformed error system is stable.The paper is organized as follows: the problem and thecontrol configuration are introduced in Section 2. Section 3presents the adaptive neural network dynamic inversionwith prescribed performance design, stability analysis, modelerror analysis, and neural network structure. And the simulations are described in Section 4. Finally, this paper concludesin Section 5.Substituting (3)-(4) into (2), and (2) can be rewritten inthe affine nonlinear form as๐๐๐ฟ๐๐ฬ[ ๐ ฬ] [ ๐๐ ] ๐บ๐ข [ ๐ฟ๐ ] ,[ ๐ฟ๐ ][ ๐ ฬ ] [ ๐๐ ]where ๐๐ , ๐๐ , ๐๐ , and ๐บ๐ข are๐๐ (๐1 ๐ ๐2 ๐) ๐ ๐3 ๐๐ฅ0 ๐4 ๐๐ง0 ,2. Aircraft Nonlinear Attitude Angle Model๐๐ ๐5 ๐๐ ๐6 (๐2 ๐2 ) ๐7 ๐๐ฆ0 ,The aircraft nonlinear attitude dynamic model can be presented as๐๐ (๐8 ๐ ๐2 ๐) ๐ ๐4 ๐๐ฅ0 ๐9 ๐๐ง0 ,๐ ฬ ๐ (๐ cos ๐ ๐ sin ๐) tan ๐,๐ฬ ๐ cos ๐ ๐ sin ๐,๐๐ฅ0 (2)๐ ฬ (๐8 ๐ ๐2 ๐) ๐ ๐4 ๐ฟ ๐9 ๐,where ๐, ๐, and ๐ are the roll, pitch, and yaw attitude angles.๐, ๐, and ๐ are the roll, pitch, and yaw angular rates. ๐1 , . . . , ๐9can be found in [17]. ๐ฟ, ๐, and ๐ are the roll, pitch, and yawmoments, which can be described as๐ฟ ๐๐ ๐2 ๐๐๐ถ๐,2๐ ๐ ๐๐ ๐2 ๐๐๐ถ๐,2(3)๐๐ ๐ ๐๐๐ถ๐,2where ๐๐ is the air density, ๐ is the wing reference area, ๐ isthe wing span, ๐ is the flight velocity, and ๐ is the wing meangeometric chord. ๐ถ๐ , ๐ถ๐ , and ๐ถ๐ are the rolling, pitching, andyawing moment coefficients described as(4)๐ถ๐ ๐ถ๐๐ฝ ๐ฝ ๐ถ๐๐ ๐ ๐ถ๐๐ ๐ ๐ถ๐๐ฟ๐ ๐ฟ๐ ๐ถ๐๐ฟ๐ ๐ฟ๐ ,๐ ๐๐/ (2๐) ,ฬ (2๐)๐ผฬ ๐ผ๐/and ๐ผฬ is the derivative of the angle of attack.,๐๐ ๐2 ๐๐ (๐ถ๐,๐ผ 0 ๐ถ๐๐ผ ๐ผ ๐ถ๐๐ ๐ ๐ถ๐๐ผฬ ๐ผ)ฬ2๐๐ ๐2 ๐๐ (๐ถ๐๐ฝ ๐ฝ ๐ถ๐๐ ๐ ๐ถ๐๐ ๐)2,(9).According to (1), we can derive the second derivatives ofattitude angles as follows:(10)where1 sin ๐ tan ๐ cos ๐ tan ๐cos ๐ sin ๐ ] ,๐ฟ (๐, ๐) [00sin๐sec๐cos๐sec๐ ][(11)ฬ ฬ๐๐sec๐ ๐๐ฬ ฬ tan ๐].ฬ ฬ cos ๐๐ (๐, ๐, ๐, ๐,ฬ ๐,ฬ ๐)ฬ [ ๐๐ฬฬฬฬ[๐๐sec๐ ๐๐ tan ๐](12)๐๐[๐,ฬ ๐,ฬ ๐]ฬ ๐ฟ (๐, ๐) [๐๐ , ๐๐ , ๐๐ ] ๐ (๐, ๐, ๐, ๐,ฬ ๐,ฬ ๐)ฬ๐where ๐ถ( ) is the aerodynamic derivatives. ๐ผ and ๐ฝ are theangles of attack and sideslip. ๐ฟ๐ , ๐ฟ๐ , and ๐ฟ๐ are the aileron, elevator, and rudder deflections, which are the control actuatorsof the aircraft. ๐, ๐, ๐, and ๐ผฬ are defined by๐ ๐๐/ (2๐)๐๐ง0 2Substituting (6) into (10), we obtain๐ถ๐ ๐ถ๐๐ฝ ๐ฝ ๐ถ๐๐ ๐ ๐ถ๐๐ ๐ ๐ถ๐๐ฟ๐ ๐ฟ๐ ๐ถ๐๐ฟ๐ ๐ฟ๐ ,๐ ๐๐/ (2๐) ,๐๐ฆ0 ๐๐ ๐2 ๐๐ (๐ถ๐๐ฝ ๐ฝ ๐ถ๐๐ ๐ ๐ถ๐๐ ๐)๐๐[๐,ฬ ๐,ฬ ๐]ฬ ๐ฟ (๐, ๐) [๐,ฬ ๐,ฬ ๐]ฬ ๐ (๐, ๐, ๐, ๐,ฬ ๐,ฬ ๐)ฬ ,2๐ถ๐ ๐ถ๐,๐ผ 0 ๐ถ๐๐ผ ๐ผ ๐ถ๐๐ ๐ ๐ถ๐๐ผฬ ๐ผฬ ๐ถ๐๐ฟ๐ ๐ฟ๐ ,(8)and๐ฬ (๐1 ๐ ๐2 ๐) ๐ ๐3 ๐ฟ ๐4 ๐,๐ ฬ ๐5 ๐๐ ๐6 (๐2 ๐2 ) ๐7 ๐,(7)๐๐ถ0๐๐ถ๐๐ฟ๐๐๐ ๐2 ๐ [๐3 0 ๐4 ] [ ๐๐ฟ๐0๐๐ถ0 ],0๐0๐บ๐ข ๐๐ฟ๐720๐๐๐๐ถ0๐๐ถ9] [๐๐ฟ๐๐๐ฟ๐ ][4(1)(๐ cos ๐ ๐ sin ๐),๐ฬ cos ๐(6)(5) ๐ฟ (๐, ๐) ๐บ๐ข [๐ฟ๐ , ๐ฟ๐ , ๐ฟ๐ ] .(13)3. Prescribed Performance-Based AdaptiveNeural Network Dynamic Inversion DesignThe aircraft attitude model shown in (13) can be representedin the following shorthand notation:๐ฅฬ ๐ (๐ฅ, ๐ฅ)ฬ ๐ (๐ฅ) ๐ข,(14)
Journal of Applied Mathematics3xฬ dxg Commandxฬ dfilterxd Controller with prescribed performancexud ctuatorxDatascalingudFigure 1: Adaptive neural network dynamic inversion with prescribed performance architecture.where the controlled state ๐ฅ [๐, ๐, ๐]๐ and the control vector ๐ข [๐ฟ๐ , ๐ฟ๐ , ๐ฟ๐ ]๐ . ๐(๐ฅ, ๐ฅ)ฬ and ๐(๐ฅ) are nonlinear functions.The state tracking error is defined as๐ (๐ก) ๐ฅ (๐ก) ๐ฅ๐ (๐ก) ,3.1. Dynamic Inversion. This section will show a brief introduction of dynamic inversion. And the readers could derivemuch more details from the reference [2].We seek to linearize a nonlinear system through computing dynamic inversion to cancel the nonlinearities in thesystem. The aircraft dynamics are shown in (14). The numberof control inputs and controlled states must be the same; thatis to say, the nonlinear function ๐(๐ฅ) is invertible. Then, thecontrol input can be calculated by(16)where ๐ข๐ is the desired response of ๐ฅ.ฬ Replacing the ๐ข in theright of (14) by the ๐ข๐ from (16), we derive๐ฅฬ ๐ข๐(17)and any nonlinearities in ๐(๐ฅ, ๐ฅ)ฬ and ๐(๐ฅ) are cancelled.The achieved system dynamics will match the chosendesired dynamics when there are no errors between thedesign model and real object. However, the model error isinevitable. So a new method is proposed to compensate themodel error and guarantee the system performances in thetransient and steady behavior.3.2. Performance Function and Error TransformationDefinition 1 (see [9]). A smooth function ๐ : R R canbe called a performance function if the following conditionsare satisfied:๐ (๐ก) 0,๐ฬ (๐ก) 0,lim ๐ (๐ก) ๐ 0.๐ก ๐ (๐ก) (๐0 ๐ ) ๐ ๐๐ก ๐ ,(19)(15)where ๐ฅ๐ (๐ก) is the desired state vector.The proposed control architecture of the aircraft attitudecontrol system is shown in Figure 1.ฬ ,๐ข๐ ๐ 1 (๐ฅ) (๐ข๐ ๐ (๐ฅ, ๐ฅ))For example, a performance function is(18)where ๐0 , ๐ and ๐ are positive constants, ๐0 is the initialtracking error ๐(๐ก), and ๐ is the maximum allowable tracking error ๐(๐ก) at the steady state. The decrement of trackingerror ๐(๐ก) will decrease when the parameter ๐ decreases. Andwe can derive the first and second derivatives of ๐(๐ก) asfollows:๐ฬ (๐ก) ๐ (๐0 ๐ ) ๐ ๐๐ก ,๐ฬ (๐ก) ๐2 (๐0 ๐ ) ๐ ๐๐ก .(20)Then by satisfying the following condition: ๐ฟ๐ (๐ก) ๐ (๐ก) ๐ฟ๐ (๐ก) , ๐ก 0,(21)where 0 ๐ฟ, and ๐ฟ 1 are prescribed scalars; the objectiveof guaranteeing transient and steady performance can bederived.Remark 2. According to (21), ๐ฟ๐(0) and ๐ฟ๐(0) are the lowerbound of the negative overshoot and upper bound of thepositive overshoot of ๐(๐ก), respectively. And a lower bound ofthe convergence speed of ๐(๐ก) is introduced by the decreasingrate of ๐(๐ก).Remark 3. By changing the parameters of performance function ๐(๐ก) and the positive prescribed scalars ๐ฟ, and ๐ฟ, themaximum overshoot and convergence rate of ๐(๐ก) can bemodified.To transform the original system with the constrainedtracking error performance (in (21)) into an equivalentconstrained one, an error transformation is introduced. Andthe error transformation is defined as๐ (๐ก) ๐ (๐ก) ๐ (๐) ,(22)
4Journal of Applied Mathematicswhere ๐ is the transformed error and a smooth and strictlyincreasing function ๐ has the following properties: ๐ฟ ๐ (๐) ๐ฟ,lim ๐ (๐) ๐ฟ,๐ฬ (23)lim ๐ (๐) ๐ฟ,๐ The derivative of (28) is๐ ๐ (0) 0.(24) ๐ฟ๐ (๐ก) ๐ (๐ก) ๐ (๐) ๐ฟ๐ (๐ก) .๐ ฬ (๐ก) ๐ (๐ก) ๐ฬ (๐ก)๐ฬ . ๐ (๐ก)๐2 (๐ก)(26)๐ฬ ๐ (๐ก) ๐ 1 (๐ (๐ก)).๐ (๐ก) ( ๐ 1 / ๐) ๐ ฬ (๐ก) ๐ (๐ก) ๐ฬ (๐ก) 2 )( ๐๐ (๐ก)๐2 (๐ก) (27)In addition, from the third property in (25),lim๐ก ๐(๐ก) 0 can be achieved if lim๐ก ๐(๐ก) 0 is satisfied.Then (22) can be described as (b) Stabilization of the transformed system using (28) issufficient to guarantee the prescribed performance.In what follows, an adaptive neural network dynamicinversion method is proposed to stabilize the transformedsystem using (28).Assumption 5. The desired states ๐ฅ๐ (๐ก) are known boundedtime functions, with known bounded derivatives.๐ ๐, ๐, ๐,(29)where ๐๐ , ๐ ๐, ๐, and ๐ are positive constants to be chosen.We define๐ (๐ก) ๐ (๐ก).๐ (๐ก)(30)(34)And the pitch and yaw errors are derived by the similarmethod.Substituting (30)โ(33) into (34), we obtain๐ธ๐ฬ ๐๐ 11๐ ฬ (๐ก) ๐ธ๐๐,๐๐ (๐ก) ๐ ๐ ๐(35)where๐ธ๐๐ ( ๐๐ 1 / ๐ ๐ ) ๐ ๐ [ ๐๐ 1 ๐ ๐(๐๐ฬ (๐ก)๐๐ (๐ก) ๐๐ (๐ก) ๐๐ฬ (๐ก)๐๐2 (๐ก)2) ๐๐ 1๐ฬ ๐ ๐ ๐2๐๐ฬ (๐ก) ๐๐ฬ (๐ก)๐๐2 (๐ก)Assumption 6. The states ๐ฅ(๐ก) of the nonlinear system in (14)are available for measurement.We define the following error function ๐ธ๐ (๐ก), whichdescribes the dynamics of the new error system using theerror transformation equation (28).๐ (๐ก) ๐ฬ (๐ก) 2๐ ฬ (๐ก) ๐2ฬ (๐ก) ].๐2 (๐ก)๐3 (๐ก)๐ธ๐ฬ (๐ก) ๐๐ฬ (๐ก) ๐๐ ๐๐ฬ (๐ก) . ๐๐3.3. Controller Design and Stability Analysis(33)Then we compute the time derivative of ๐ธ๐ (๐ก) for the rollerror as(a) The system in (14) is invariant under the error transformation equation (22).๐ธ๐ (๐ก) ๐๐ฬ (๐ก) ๐๐ ๐๐ (๐ก) , ๐ 1 ๐ ฬ (๐ก) 2๐ ฬ (๐ก) ๐ฬ (๐ก)[ ๐ ๐ (๐ก)๐2 (๐ก)(28)Lemma 4 (see [9]). Consider system in (14), the transientand steady state tracking error behavior bounds described bythe performance function ๐(๐ก) and the error transformationequation (22). The following results hold.(32)And the second derivative of (28) isAccording to (19), we obtain ๐ฟ๐ (๐ก) ๐ (๐ก) ๐ฟ๐ (๐ก) .(31)where(25)According to the first property in (23) and ๐(๐ก) 0, wehave ๐ 1 ฬ๐, ๐ ๐๐ (๐ก) ๐๐ฬ (๐ก)๐๐2 (๐ก) 2๐๐ฬ (๐ก) ๐๐2ฬ (๐ก)๐๐3 (๐ก)].(36)Then we can derive๐ธฬ ๐ธ๐ ๐ ฬ (๐ก) ๐ธ๐,(37)๐ฬ [๐๐ฬ (๐ก), ๐๐ฬ (๐ก), ๐๐ฬ (๐ก)]๐ , ๐ธ๐ where ๐ธฬ [๐ธ๐ฬ , ๐ธ๐ฬ , ๐ธ๐ฬ ] , ๐(๐ก)๐[๐ธ๐๐, ๐ธ๐๐, ๐ธ๐๐] , and ๐ธ๐ is๐ธ๐๐ [๐ธ๐ [๐ธ๐๐ []],๐ธ๐๐ ](38)
Journal of Applied Mathematics5where ๐ธ๐๐ ( ๐๐ 1 / ๐ ๐ )/๐๐ (๐ก), ๐ธ๐๐ ( ๐๐ 1 / ๐ ๐ )/๐๐ (๐ก),๐ธ๐๐ ( ๐๐ 1 / ๐ ๐ )/๐๐ (๐ก), and ๐๐ฬ (๐ก), ๐๐ฬ (๐ก), ๐๐ฬ (๐ก), are๐๐ฬ (๐ก) ๐ ฬ ๐๐ฬ ,๐๐ฬ (๐ก) ๐ฬ ๐๐ฬ ,(39)The control input of roll channel is 1๐ข๐ ๐ฟ๐ ๐บ๐ 1 [ ๐น๐ (๐ธ๐๐ ) (๐ธ๐๐ ๐๐ ๐ธ๐ ) ๐๐ฬ ๐ข๐๐๐ ] .(43)The adaptive signal of roll channel is๐๐ฬ (๐ก) ๐ฬ ๐ฬ๐ .๐ข๐๐๐ ๐ค๐๐ ๐๐ .To simplify the controller design progress, we linearize (2)in an equilibrium point which is the steady wings-level flightstate.The neural network weight update law of roll channel is๐[๐,ฬ ๐,ฬ ๐]ฬ ๐ด ๐ [๐0 ฮ๐, ๐ผ0 ฮ๐ผ, ๐ฝ0 ฮ๐ฝ, ๐0๐ ฮ๐, ๐0 ฮ๐, ๐0 ฮ๐](40)๐ ๐ต๐ [๐ฟ๐0 ฮ๐ฟ๐ , ๐ฟ๐0 ฮ๐ฟ๐ , ๐ฟ๐0 ฮ๐ฟ๐ ] ,where the ๐ด ๐ and ๐ต๐ are the appropriate dimension constantmatrixes, ๐ฝ0 ๐0 ๐0 ๐0 ๐ฟ๐0 ๐ฟ๐0 0. And๐0 , ๐ผ0 , and ๐ฟ๐0 are the flight velocity, angle of attack andelevator deflection in some equilibrium point, respectively.The symbol ฮ represents the small perturbation from theequilibrium value.According to (2), (13), (14), and (40), we can obtain๐ฅฬ ๐น (๐ฅ) ๐บ (๐ฅ) ฮ๐ข ๐,(41)๐๐พ (๐ (๐ธ ) ๐ธ๐๐ ๐๐ ๐ค๐ ) ,ฬฬ ๐ { ๐ ๐ ๐๐ค0(44) ๐ธ๐ ๐๐ , ๐ธ ๐ , ๐ ๐22 ๐ ๐ธ๐ โ๐ ( ๐ธ๐๐ โ๐ ) ๐๐ ๐๐ (๐ค๐max ) ๐๐ ,2๐๐where the vector ๐๐ is a set of basis functions to approximatethe uncertainty and the neural network weight vector ๐ค๐ isthe set of coefficients of each basis function in the roll channel.The adaptation gain ๐พ๐ determines the learning rate of neuralnetwork. The ๐๐ is a modification term to limit the growth ofthe neural network weights. The constant ๐๐ is positive. And thepositive constant โ๐ is the neural network approximate errorwhich is bounded. The neural network weight error isฬ๐ ๐ค๐ ๐ค๐ ,๐ค๐where ฮ๐ข [ฮ๐ฟ๐ , ๐ฟ๐0 ฮ๐ฟ๐ , ฮ๐ฟ๐ ] , and(45)(46)where the ๐ค๐ is the true value of the neural network weight inthe roll channel.๐๐น (๐ฅ) [๐น๐ , ๐น๐ , ๐น๐ ] ๐ (๐, ๐, ๐, ๐,ฬ ๐,ฬ ๐)ฬ๐ ๐ฟ (๐, ๐) ๐ด ๐ [๐0 ฮ๐, ๐ผ0 ฮ๐ผ, ฮ๐ฝ, ฮ๐, ฮ๐, ฮ๐] ,๐๐บ (๐ฅ) [๐บ๐ , ๐บ๐ , ๐บ๐ ] ๐ฟ (๐, ๐) ๐ต๐๐๐ [๐๐ , ๐๐ , ๐๐ ] ,(42)where ๐ is the model error which will be analyzed inSection 3.4.The formula ๐ฅฬ ๐น(๐ฅ) ๐บ(๐ฅ)ฮ๐ข in (41) is namedas the design model in some equilibrium point, which isdifferent from the real nonlinear model in (14). And thedifference between the design model and the nonlinear modelis represented by the symbol ๐, which will be compensated bythe adaptive neural network.Because there are three channels in the attitude controland the form of each channel is the same, consider thefollowing Theorem 7 for the roll channel. And the pitch andyaw channels are similar.Theorem 7. Considering Assumption 5, Assumption 6, andthe nonlinear system in (14), all the signals are bounded, andthe tracking error ๐(๐ก) satisfies the performance described bythe performance function ๐(๐ก), if the control input of systemsatisfies the following formula.Proof. A suitable Lyapunov function of roll channel will be๐๐11ฬ๐ ,{(ฬ๐ค๐ ) ๐ค(๐ธ ) ๐ธ๐ {{2 ๐2๐พ๐๐๐ { 1๐๐1{{ (๐ธ๐0 ) ๐ธ๐0 ฬ๐ ,(ฬ๐ค๐ ) ๐ค2๐พ๐{2 ๐ธ๐ ๐๐ , ๐ธ๐ ๐๐ , (47)where โ๐ธ๐0 โ ๐๐ and ๐๐ is to be defined later.Firstly, if โ๐ธ๐ โ ๐๐ is satisfied, then the time derivative of(47) is given by๐๐1ฬฬ ๐ .๐๐ฬ (๐ธ๐ ) ๐ธ๐ฬ (ฬ๐ค๐ ) ๐ค๐พ๐(48)Substituting (37) into (48), we derive๐๐1ฬฬ ๐ .๐๐ฬ (๐ธ๐ ) [๐ธ๐๐ ๐ธ๐๐ (๐ ฬ ๐๐ฬ )] (ฬ๐ค ) ๐ค๐พ๐ ๐(49)Considering (41)-(42) and (49), we have๐๐๐ฬ (๐ธ๐ ) [๐ธ๐๐ ๐ธ๐๐ (๐น๐ ๐บ๐ ๐ข๐ ๐๐ ๐๐ฬ )] ๐1ฬฬ ๐ .(ฬ๐ค๐ ) ๐ค๐พ๐(50)
6Journal of Applied MathematicsLet the control input ๐ข๐ satisfy (43), then (50) can bedescribed as๐๐1ฬฬ ๐ . (51)๐๐ฬ (๐ธ๐ ) [ ๐๐ ๐ธ๐ ๐ธ๐๐ (๐๐ ๐ข๐๐๐ )] (ฬ๐ค ) ๐ค๐พ๐ ๐Substituting (44)โ(46) into (51), we obtain๐๐๐๐ฬ ๐๐ (๐ธ๐ ) ๐ธ๐ (๐ธ๐ ) ๐ธ๐๐ ๐(52)๐ (๐๐ (๐ค๐ ) ๐๐ ) ๐๐ (ฬ๐ค๐ ) ๐ค๐ .By using the norms of the terms on the right side of (52),we obtain the following inequality: 2 ๐๐ฬ ๐๐ ๐ธ๐ ๐ธ๐ ๐ธ๐๐ ฬฬ ๐ 0, andHere the weight update law is ๐คฬ ๐ ๐ค๐๐ฬ 0. Therefore, the system is stable, and all the signalsare bounded. Considering Lemma 4, the tracking error ๐(๐ก)satisfied the performance described by the performancefunction ๐(๐ก).This completes the proof.3.4. Analysis of the Model Error. According to (2)โ(5), themoment model is nonlinear, complicated, and must becontinuously varying with the flight condition. For simplicity,the linear model of (40) in an equilibrium point is used toreplace the nonlinear model of (2).We define the model error ฮ [ฮ ๐ , ฮ ๐ , ฮ ๐ ]๐ , whichis the error between the linear model Equation (40) and thenonlinear model equation (6). And the ฮ is(53) ๐ ฬ๐ ๐ค๐ . ๐๐ (๐ค๐ ) ๐๐ ๐๐ ๐ค ๐ฮ [๐๐ , ๐๐ , ๐๐ ] ๐ด ๐ [๐0 ฮ๐, ๐ผ0 ฮ๐ผ, ฮ๐ฝ, ฮ๐, ฮ๐, ฮ๐]In addition, the approximate error of neural network isbounded, so the following equation is satisfied: ๐๐ (๐ค )๐ ๐๐ โ๐ .๐ (54)The maximum weight of ideal neural network in the rollchannel is ๐ค๐max , so we have ๐ค๐ ๐ค๐max . (55)๐ 2 2๐๐ฬ ๐๐ ๐ธ๐ ๐ธ๐ ๐ธ๐๐ โ๐ ๐๐ (๐ค๐max ๐ค๐ ๐ค๐ ) .(56)(61)Then (6) can be rewritten as๐[๐,ฬ ๐,ฬ ๐]ฬ ๐ด ๐ [๐0 ฮ๐, ๐ผ0 ฮ๐ผ, ฮ๐ฝ, ฮ๐, ฮ๐, ฮ๐]๐๐ฅฬ ๐น (๐ฅ) ๐บ (๐ฅ) ฮ๐ข ๐ฟ (๐, ๐) ฮ.2) ๐๐ (๐ค๐max22(57)).If the system is stable, then ๐๐ฬ 0. And (57) can betransformed to the following formula:max2๐ค๐ 2 ๐๐ ๐ธ๐ ๐ธ๐ ๐ธ๐๐ โ๐ ๐๐ () 0.2(58)Then we can derive22 ๐ ๐ธ โ ( ๐ธ๐๐ โ๐ ) ๐๐ ๐๐ (๐ค๐max ) ๐ ๐(59) ๐ธ๐ ๐๐ . 2๐๐Next, if โ๐ธ๐ โ ๐๐ is satisfied, then the time derivative of(47) is derived by๐1ฬฬ ๐ .๐๐ฬ (ฬ๐ค๐ ) ๐ค๐พ๐(62)(64)Therefore, the model error mainly depends on the different equilibrium points, attitude angles, actuator deflections,and so on.3.5. Neural Network Structure. The first step in determiningthe appropriate network structure is identifying the networkinputs. Based on the analysis of model error sources describedin Section 3.3, there are three main categories of inputs: theattitude angles, attitude angle rates, and actuator deflections.A Sigma-Pi neural network [18] is used to compensate themodel error ๐, and the basis function of the pitch channel ๐๐is๐๐ kron (kron (๐ถ1๐ , ๐ถ2๐ ) , ๐ถ3๐ ) ,(65)where kron( , ) represents the Kronecker products and isdefined as follows:2 ๐(60)(63)Comparing (63) to (41), we obtain๐ ๐ฟ (๐, ๐) ฮ. 2 ๐๐ฬ ๐๐ ๐ธ๐ ๐ธ๐ ๐ธ๐๐ โ๐2๐ ๐ต๐ [ฮ๐ฟ๐ , ๐ฟ๐0 ฮ๐ฟ๐ , ฮ๐ฟ๐ ] ฮ.Considering (56), we obtain ๐๐ ( ๐ค๐ ๐ ๐บ๐ข [๐ฟ๐ , ๐ฟ๐ , ๐ฟ๐ ] ๐ต๐ [ฮ๐ฟ๐ , ๐ฟ๐0 ฮ๐ฟ๐ , ฮ๐ฟ๐ ] .Substituting (62) into (10), we haveSubstituting (46) and (54)-(55) into (53), we get๐ค๐max๐๐ถ1๐ [1, ๐, ๐ ] ,๐๐ถ2๐ [1, ๐] ,2 ๐๐ถ3๐ [1, ๐ฟ๐ , ๐ฟ๐ ] ,(66)
Journal of Applied Mathematics7๐ฬ d1ฮ ฬqw๐๐๐g2๐ฬ d๐๐2 1/s 1/s๐dฮ 1ฮฃ.q2๐๐ ๐๐uqad๐๐21Figure 3: Command filter.๐ฟeฮ 2๐ฟeTable 1: Performance parameters.Figure 2: Neural network structure.๐where ๐, ๐, ๐ฟ๐ and ๐ฟ๐ are normalized variables between 1 and1. The normalization function is๐ฆ ๐ (๐ฅ) 2 1,1 ๐ 0.1๐ฅ(67)where ๐ฅ is the input parameter and ๐ฆ is the output parameter.And a general description of the neural network is shownin Figure 2.And the basis function of roll channel ๐๐ and the basisfunction of yaw channel ๐๐ can be derived similarly as follows:๐๐ kron (kron (kron (kron (๐ถ1 , ๐ถ2 ) , ๐ถ3 ) , ๐ถ4 ) , ๐ถ๐ ) , (68)๐๐0๐๐ ๐๐๐ฟ๐ 12 0.3 0.70.6๐0๐๐ ๐๐๐ฟ๐ 8 0.2 0.70.5๐0๐๐๐ ๐๐๐ฟ๐ 10 0.2 0.70.6๐ฟ๐1๐ฟ๐1๐ฟ๐1Table 2: Controller 0500.3where ๐ ๐, ๐. Then ๐ถ๐ , ๐ 1, 2, 3, 4, ๐ is2 ๐๐ถ1 [1, ๐, ๐ ] ,2 ๐๐ถ4 [1, ๐, ๐ ] ,๐๐ถ2 [1, ๐] ,๐ถ3 [1, ๐]๐ ,2 ๐๐ถ๐ [1, ๐ฟ๐ , ๐ฟ๐ ] ,2 ๐๐ถ๐ [1, ๐ฟ๐ , ๐ฟ๐ ] .(69)4. Simulation ResultsIn this section, we consider the attitude angles controlproblem for a fixed-wing aircraft, and the initial flight stateis the wings-level flight. Then the attitude angles commandsin three channels will be tracked, respectively.In the following simulation, the initial flight height andvelocity are 6000 m and 190 m/s, and the initial attitude anglesand angular rates including ๐, ๐, ๐, ๐, ๐, and ๐ are zeros. Inaddition, all the initial actuator deflections are zeros.The error transformation function [19] in the simulationis described as๐ (๐) ๐ฟ๐(๐ ๐ฆ) ๐ฟ๐ (๐ ๐ฆ),๐(๐ ๐ฆ) ๐ (๐ ๐ฆ)(70)where ๐ฆ ln(๐ฟ/๐ฟ)/2. It can be shown that ๐(๐) satisfies theproperties in (23)โ(25).The attitude angles commands of three channels aretransformed into the desired attitude angles commandsthrough the command filters. And the structure of commandfilter for the roll channel is shown in Figure 3. In addition, thedesired attitude angles commands for yaw and pitch channelscan be obtained by the similar command filters.The command filter parameters are set as ๐๐ 1, ๐๐ 2.5,and ๐ ๐, ๐, ๐.Design the control inputs with prescribed performancefor three channels through the procedures in Section 3.2. Theperformance and controller parameters are shown in Tables 1and 2.Remark 8. For the performance function ๐(๐ก), ๐0 is derivedby subtracting the attitude command from the initial attitudeangle. ๐ is the allowable attitude tracking error at the steadystate. And the decrement of tracking error ๐(๐ก) will decreasewhen the parameter ๐ decreases.Remark 9. For the controller parameters, the adaptation gain๐พ will improve the attitude tracking performance, especially,when there are much larger model errors. The ๐๐ is amodification term to limit the growth of the neural networkweights; therefore, it is small. The transient performanceof attitude tracking error can be improved by increasingthe parameter ๐; however, the increase will increase themagnitude of the control input. Then a compromise must bereached.
8Journal of Applied Mathematics15Yaw angle (deg)Roll angle (deg)151050 50102030Time (s)40501050 5600102030405060Time (s)(b)Pitch angle (deg)(a)151050 50102030Time (s)405060Attitude angle commandResponse of method [20]Response of proposed method(c)Figure 4: Responses of the attitude angles.10Yaw error (deg)Roll error (deg)20100 10 200102030Time (s)40506050 5 100102030Time (s)Pitch error (deg)(a)405060(b)1050 5 100102030Time (s)Error of method [20]Error of proposed method405060The upper bound of errorThe lower bound of error(c)Figure 5: Tracking errors of the attitude angles.The design model I is derived at the trimmed flightcondition of 6000 m and 190 m/s, and the model error issmall.The aircraft tracks the attitude angles commands from theinitial flight state. And the attitude angles tracking responsesand tracking errors are shown in Figures 4 and 5.The two methods have achieved the attitude angles command tracking. Figure 4 shows the better attitude responsesare achieved by the proposed method compared to themethod in [20]. And the coupling among different channelsis smaller when the proposed method is used. For example,the roll angle response has a less change when the aircrafttracks the yaw command. In Figure 5, the attitude anglestracking errors satisfy the prescribed performance boundwith the proposed method in the dynamic and steady state.The main reason is that the method in [20] does not considerthe performance bound defined by the performance function๐(๐ก) in the design process.
Journal of Applied Mathematics910Yaw angle (deg)Roll angle (deg)151050 50102030Time (s)405050 560010Roll commandRoll response2030Time (s)405060Yaw commandYaw response(a)(b)Pitch angle (deg)1050 50102030Time (s)405060Pitch commandPitch response(c)10105Yaw error (deg)200 10 200102030Time (s)40500 5 1060010Roll tracking error๐p (t) 0.6๐p (t)2030Time (s)Yaw tracking error๐r (t) 0.5๐r (t)(a)(b)10Pitch error (deg)Roll error (deg)Figure 6: Responses of the Attitude angles with model error.50 5 100102030405060Time (s)Pitching tracking error๐q (t) 0.6๐q (t)(c)Figure 7: Tracking errors of the attitude angles with model error.405060
10Journal of Applied Mathematics105Rudder (deg)Aileron (deg)100 5 10010203040500 10 20600102030Time (s)405060405060Time (s)(a)(b)Elevator (deg)100 10 200102030405060Time (s)Design model IDesign model II(c)Figure 8: Deflections of the control actuators in two design models.Yaw channel uad50 20 400102030Time (s)40500 5 106001020(a)30Time (s)(b)20Pitch channel uadRoll channel uad200 20 40 600102030Time (s)4050Design model IDesign model II(c)Figure 9: The outputs of neural network in three channels.60
Journal of Applied MathematicsIn the real flight control system, there must be the modelerror. In order to verify that the similar tracking performanceis also achieved when there is the large model error, we haveconducted the following simulation study.The flight condition is the same, and the initial flightheight and velocity are 6000 m and 190 m/s. However, thedesign model II used to design the attitude angles controllersis derived at the trimmed flight condition of 4000 m and150 m/s. Apparently, the model error is large.And the attitude angles tracking responses and trackingerrors are shown in Figures 6 and 7.Figures 6 and 7 show the attitude angles tracking errorsstill satisfy the prescribed performance bound, although themodel error is large in this situation. In addition, Figures6 and 7 show the track performance is similar when thedifferent design models are used.The control actuators deflections for three channels arecompared in Figure 8 when the two design models are used.Figure 8 shows the deflections of the control actuatorsusing the design model II increase to derive the desiredattitude angles tracking performance. In addition, the outputsof neural network in three channels are shown in Figure 9.Figure 9 shows the outputs of neural network using thedesign model II are larger than the one using the designmodel I. The main reason is that the model error is largerwhen the design model II is used, and the larger outputs ofneural network are used to compensate the large model error.5. ConclusionIn this paper, an adaptive neural network dynamic inversion with prescribed performance method is proposed foraircraft attitude control. By incorporating the adaptive neuralnetwork dynamic inversion with the prescribed performanceconcept, the proposed method guarantees the system tracking error satisfies the prescribed performance bound in thetransient and steady behavior. The nonlinear simulation ofthe aircraft also verifies the effectiveness of the proposedapproach.Further investigation is needed for the situations in thepresence of the external wind disturbance and unmodeleddynamics. And, these design parameters in this methodshould be decreased and optimized to achieve a real application.AcknowledgmentsThis work was supported by the Program for New CenturyExcellent Talents in University (Grant no. NCET-10-0032)and by the National Natural Science Foundation of China(Grant no. 61175084).References[1] S. A. Snell, D. F. Enns, and W. L. Garrard Jr., โNonlinear inversion flight control for a supermaneuverable aircraft,โ Journal ofGuidance, Control, and Dynamics, vol. 15, no. 4, pp. 976โ984,1992.11[2] D. Enns, D. Bugajski, R. Hendrick, and G. Stein, โDynamicinversion: an evolving methodology for flight control design,โInternational Journal of Control, vol. 59, no. 1, pp. 71โ91, 1994.[3] J. Leitner, A. Calise, and J. V. R. Prasad, โAnalysis of adaptiveneural networks for helicopter flight control,โ Journal of Guidance, Control, and Dynamics, vol. 20, no. 5, pp. 972โ979, 1997.[4] R. Rysdyk and A. J. Calise, โRobust nonlinear adaptive flightcontrol for consistent handling qualities,โ IEEE Transactions onControl Systems Technology, vol. 13, no. 6, pp. 896โ910, 2005.[5] C. Cao and N. Hovakimyan, โDesign and analysis of a novel๐ฟ 1 adaptive control architecture with guaranteed transientperformance,โ IEEE Transactions on Automatic Control, vol. 53,no. 2, pp. 586โ591, 2008.[6] W. Gai, H. Wang, and D. Li, โTrajectory tracking for automatedaerial refueling based on adaptive dynamic inversion,โ Journalof Beijing University of Aeronautics and Astronautics, vol. 38, no.5, pp. 585โ590, 2012 (Chinese).[7] H. Xu and P. A. Ioannou, โRobust adaptive control for a class ofMIMO nonlinear systems with guaranteed error bounds,โ IEEETransactions on Automatic Control, vol. 48, no. 5, pp. 728โ742,2003.[8] V. Stepanyan and K. Krishnakumar, โAdaptive control withreference model modification,โ Journal of Guidance, Control,and Dynamics, vol. 35, no. 4, pp. 1370โ1374, 2012.[9] C. P. Bechlioulis and G. A. Rovithakis, โAdaptive control withguaranteed transient and steady state tracking error bounds forstrict feedback systems,โ Automatica, vol. 45, no. 2, pp. 532โ538,2009.[10] C. P. Bechlioulis and G. A. Rovithakis, โPrescribed performanceadaptive control for multi-input multi-output affine in thecontrol nonlinear systems,โ IEEE Transactions on AutomaticControl, vol. 55, no. 5, pp. 1220โ1226, 2010.[11] A. Kostarigka and G. Rovithakis, โAdaptive dynamic outputfeedback neural network control of uncertain MIMO nonlinearsystems with prescribed performance,โ IEEE Transactions onNeural Networks and Learning Systems, vol. 23, no. 1, pp. 138โ149, 2012.[12] C. P. Bechlioulis, Z. Doulgeri, and G. A. Rovithakis, โGuaranteeing prescribed performance and contact maintenance via anapproximation free robot force
by dynamic model inversion is well known and has been applied to the control of high angle of attack ghter aircra [ , ]. e primary drawback of dynamic inversion for aircra ight control is the need for high- delity nonlinear model which must be inverted in real time. However, it is di cult to obtain the exact aircra dynamic model in practice. e .
Amendments to the Louisiana Constitution of 1974 Article I Article II Article III Article IV Article V Article VI Article VII Article VIII Article IX Article X Article XI Article XII Article XIII Article XIV Article I: Declaration of Rights Election Ballot # Author Bill/Act # Amendment Sec. Votes for % For Votes Against %
Performance comparison of adaptive shrinkage convolution neural network and conven-tional convolutional network. Model AUC ACC F1-Score 3-layer convolutional neural network 97.26% 92.57% 94.76% 6-layer convolutional neural network 98.74% 95.15% 95.61% 3-layer adaptive shrinkage convolution neural network 99.23% 95.28% 96.29% 4.5.2.
Adaptive Control, Self Tuning Regulator, System Identification, Neural Network, Neuro Control 1. Introduction The purpose of adaptive controllers is to adapt control law parameters of control law to the changes of the controlled system. Many types of adaptive controllers are known. In [1] the adaptive self-tuning LQ controller is described.
Sybase Adaptive Server Enterprise 11.9.x-12.5. DOCUMENT ID: 39995-01-1250-01 LAST REVISED: May 2002 . Adaptive Server Enterprise, Adaptive Server Enterprise Monitor, Adaptive Server Enterprise Replication, Adaptive Server Everywhere, Adaptive Se
A growing success of Artificial Neural Networks in the research field of Autonomous Driving, such as the ALVINN (Autonomous Land Vehicle in a Neural . From CMU, the ALVINN [6] (autonomous land vehicle in a neural . fluidity of neural networks permits 3.2.a portion of the neural network to be transplanted through Transfer Learning [12], and .
neural networks and substantial trials of experiments to design e ective neural network structures. Thus we believe that the design of neural network structure needs a uni ed guidance. This paper serves as a preliminary trial towards this goal. 1.1. Related Work There has been extensive work on the neural network structure design. Generic algorithm (Scha er et al.,1992;Lam et al.,2003) based .
Different neural network structures can be constructed by using different types of neurons and by connecting them differently. B. Concept of a Neural Network Model Let n and m represent the number of input and output neurons of a neural network. Let x be an n-vector containing the external inputs to the neural network, y be an m-vector
2. Neural Network in Nonlinear System Identification and Control . In the identification stage of the adaptive control of nonlinear dynamical system, a neural network identifier model for the system to be controlled is developed. Then, this identifier is used to represent the system while train-ing the neural network controller weights in the .