International Conference on Mechatronics and Automation
August 9 - 12, Changchun, China Dynamic Surface Backstepping Control Design for One Hypersonic Vehicle Chen lie Department ofControl Engineering Naval Aeronautical andAstronautical University Yantai,Shandong Province, China hjhychenjie@tom.com Wu Jinhua Department ofControl Engineering Naval Aeronautical andAstronautical University Yantai,Shandong Province, China hjhychenjie@tom.com Pan Changpeng Department ofCommand Naval Aeronautical and Astronautical University Yantai,Shandong Province, China navypcp@ tom.com (2) (1) Abstract - Based on multilayer neural networks, feedback linearization technology, and backstepping design method, a novel robust adaptive control design method is proposed for one hypersonic vehicle(HSV) uncertain MIMO nonaffine block control system. Multilayer neural networks are used to identify the nonlinear uncertainties of the system. A continuous robust term is adopt to minify the influence of the multilayer neural networks construction error, the dynamic surface control strategy to eliminate "the explosion of terms" by introducing a series of first order filters to obtain the differentiation of the virtual control inputs. Finally, nonlinear six-degree-of-freedom (6-DOF) numerical simulation results for a HSV model are presented to demonstrate the effectiveness of the proposed method. Index Terms -hypersonic vehicle ,. neural networks ,. dynamic surface ,. backstepping of a 6-DOF HSV model is preformed to verify the effectiveness of the proposed algorithm and the conclusions are given. II. NONLINEAR HSV MODEL WITH UNCERTAINTIES Winged-cone'!' is one of the main investigate objects with its hypersonic mach number and open aerodynamic parameters, but most research is based on its longitudinal channel model. In this paper we finished Winged-cone's three-axes, highly nonlinear model with general set of uncertainties. The nonlinear dynamic equations are given as follows(the full HSV model is given in appendix, and parameters are obtained from literature [1]): Xl T, (xl'x 2 ) x 2 == h (X 2)+ gI (X 2)X3 + WI (x 2)u We assume that .h ' gI , hI ' J. ' f3 and g2 are unknown and h == hO +4h, gI == gIO + 'h == g2 ==g20 =h lO + where flO , gIO , w lO , f20 , g20 are nominal system parameters, the others are uncertainties. The meaning of the symbols used in this paper can be founded in appendix and literature[ 1]. The task of the controller is to track the commands signals when aerodynamic model uncertainties exist. Assumption 1: Ignoring the effect of the fin deflection on the aerodynamic force and viewing it as a part of the uncertainties, namely WI (Xl) == 0. where, (3) q f3 x -r X 2 = [a 6 a 6 r ]T x -[v 1- I. INTRODUCTION Hypersoninc vehicles Control technology has attracted a lot of attention in recent years because of highly nonlinear dynamics and high Mach velocity[1,2,3]. Wang[4] adopted adaptive sliding controller to analyze the longitudinal dynamics of a generic hypersonic air vehicle. Shtessel'F' designed inner and outer sliding mode controller loop to investigate the reentry of hypersonic vehiles, Austin[6] use fuzzy logic control to compensate the unknown nonlinearities, genetic algorithm to optimize the scaling parameters of fuzzy controller. a kind of neural adaptive controller for hypersonic vehicle is proposed by Xu, Mirmirani and Ioannou [7] A few recent contributions have also attempted design directly on nonlinear models [8-10]. Most research results were restricted to linearly parametrized and state feedback linearizable. Different from above work, The paper investigate a MIMO nonaffine nonlinear hypersonic vehicle model with general set of uncertainties. Dynamic surface backstepping controller is designed based on Radial-Basis-Function (RBF) neural networks approach. Stability analysis is also shown by the Lyapunov stability theory. Finally, a numerical simulation 978-1-4244-2693-5/09/$25.00 2009 IEEE 4770 III. CONTROLLERDESIGN A. RBF Neural network In the controller design RBF NN is used to approximate the unknown function rjJ(o) . In a compact set, RBF NN can be expressed as (-) =WS T (-) (4) Where WE RPis the tuning weight matrix and rjJ(-) is a vector composed of Gauss functions. Given a smooth function rjJ(-) , where n is a compact subset of Rm , and 5 > 0 , there exist a Gauss function vector S : R" H RP and an optimal weight matrix W* such that I(x) - W' ST(x)1 :-:::; l5, Vx En. Define (10) X 2d O is the certain part of x;, Multilayer NN: *T (*T ) SI XI +1 can be used to approach the uncertain part of X; ,where XI is the input of above NN, I1/ <Iu is the NN approaching error, and Iu >0 . Then we choose desired virtual control value as follow: 0< giO <Igi 8f )/8x i + 1 1 <OO''VX n En,i=l,ooo,n-l as NN reconstruction error. The optimal weight matrix W* is only used in the analysis. Assumption 2: There exists constant s; > s;> 0 to make
h(X)-W*ST(X) I1h(x) (5)
Where II > 0 is designed parameter, 'i is the inducted robust item and will be embodied as follow. W; is the estimate of is the estimated value of .Defining estimation error as =W; - , = - , then the dynamic function of error state value ZI ' then B. Dynamic surface backstepping controller design Step 1: Define error surface Zi f1J the system anticipant track, ZI derivative is 2"1 = +glA, (Z2 +Xz - -Sd)= +glA, (Z2 -llzl) +glA, +d u1 - &1 +1j ] (6) Suppose Defining VI =-X l d +klz l, k l > O. By using assumption 2 and 8v I /8x 2 =0 , we can get the following inequality a[J;(Xl'X 2)+VI]/aX2 >glo >0 From literature [11] we can get that there has an ideal virtual control function x 2 = x; (XI' VI ) , so Suppose 1 2 1 - T -I - 1 {-T -I -} =--ZI r +-tr (12) 2g l A 2 2 1 r WI = > 0, r VI = > 0 are parameters to be designed, Taking the time derivative of ,we have =rWI [ -(SI XI )ZI ] lit >a (14) = r VI [ S;ZI ] According to function (6),(7),so that 2"1 = -klz l +glA, (x 2 - x;) (9) 4771 Step 2: Define error state value as Z2 =x 2 - X 2d ' X 2d can be obtained by the following first-order filter by state variable x 2 (16) So By the error surfaces Z2 =x 2 - x 2d So we have 2 2 =flO (.\2) +glO(.\2).\3 -iy+(I:f;. (.\2) +4?, (.\2).\3) _ () () . (15) - ho x 2 +glO x 2 x 3 -x 2d - L1 where L1 = -(L1J; (x 2 ) +L1g, (X 2)X3) is the uncertain term. By employing the neural networks to approximate L1. We choose desired virtual control value as follow: -U;S2(C' ;X 2 )-rz ] Candidate Lyapunov function and adaptive laws and the robust term are selected similar as step I. Step 3: Define the error surfaces Z3 =x 3 - X 3d . In this step, we consider the third equation of (8) to design the actual control input for the hypersonic aircraft. Let X 3d pass through a first-order filter, a new state variable X 3d can be obtained IV. N UMERICAL SIMULAnON RESULTS In order to validate the effectiveness and validity of the proposed method, nonlinear 6-DOF simulation results for a HSV model are presented. The controller design parameters are chosen as follows: II =11,/ 2 =13,/ 3 =7.3 ' k =20 , (Tw, =(Tv; =0.01 ' 7]; =1,i =1,2,3 ,and X l d =[5130 2 2]. Simulation results are shown in the following figures. ,: / " ( " 0\ E ' 00 1 2 3 ! S 6 1 8 9 10 t (s) (17) Fig.1 Tracking curve of r After differentiating the error surface Z3' we get We choose the control input U = (!3 Z3 +ho - X 3d ) + ) +kz 3 + 8 3 V; X 3 +r 3 (18) (19) .QC60 1 2 3 ! 5 6 1 8 9 10 t (s) Fig.2 Tracking curve of P 1 3 , k > 0 is design parameters. Choose the following adaptive laws and the robust term W; = r W3 [(S3 )z; ] =r V3 [ ] +2) /7B (20) Suppose 4772 0\ '
00 1 2 3 4 5 6 1 8 9 10 t(s) Fig.3 Tracking curve of a 4 , 10-3 :v 5 6 7 8 9 10 t( s) FigA Tracking curve of fJ 1 [qSC La +Tsina-mgcosycos,u] mvcos fJ . _l_[qSCypfJ- Tcosa sin fJ+ mg cos ysin,u] mv ' 1;3 / 13 = _l_[qSC L a(tan y sin zz + tan13) - mgcos y cosu tan13 mv ' o mv qSC y O , ,. mv 1 sin atgfJ 1 o cosa o sinasecfJ qSCL,oa mvcosfJ qSCy,oa mv +qSCy,pf3tany cos,ucosf3+T(sina tanys in,u +sina tan13 -cos a tan y cos,usin13)] r - cos a tg fJ gl ( XI ) = sin a cosasecfJ qSCL,o, mvcos fJ qSCy,o, 10 1 '1 3 4 5 6 7 8 9 10 t(s) o '()2 0 1 2 3 4 5 6 7 8 9 10 t(s) Fig6 Tracking curve orr of\;1 \---- - - - --1 Fig.S Tracking curve ofq . 150 1 '1 3 4 5 6 7 8 9 10 t(s) Fig.7 tuning weight From the simulation results, it can be shown that the control system has good stability, performance and robustness, even if the large model uncertainties exist. V .CONCLUSIONS This paper presented a nonlinear adaptive controller for a HSV with a general set of uncertainties. The nonlinear adaptive controller is designed using dynamic surface backstepping control techniques and fully tuned RBF neural networks. Lyapunov stability theory is used to prove the stabil ity of the system and derive the tuning rules for updating all parameters of the RBF neural networks. Finally, a nonlinear 6-DOF numerical simulation of a HSV model is performed to demonstrate the good performance of the proposed method. 1 . fJ W 31 =-[qSC u (tan ysm,u + tan ) mv " +qSCy,o, tan r cos ,ucos fJ] W 32 =_I_[qSC L . o (tanysin,u+tanfJ) mv a +qSCy,oatan r cos u cos fJ] W 33 = _1_[qSC L o. (tan ysin,u + tan fJ) mv " +qSCy0 tan y cos u cos fJ] .,. APPENDIX The winged-cone dynamics model 4773 qbSG,6 a qbSG,4 I p I p I p g2= g21 g22 g23 +Xcgc:;,,6 e ) +Xcgc:;,,6 a) +xcgc:;,,4) If" 1,. 1,. REFERENCES [1] J. D.Shaughnessy, S. Z. Pinckney, J. D. Mcminn, C. I. Cruz and M. L. Kelley. Hypersonic Vehicle Simulation Model: Winged-Cone Configuration. NASA TM-I02610,1990. [2] I. M. Gregory, R. S. Chowdhry, J. D. McMinn and 1. D. Shaughnessy. Hypersonic vehicle model and control law development using Hcoand J.!- synthesis. NASA TM-4562, 1994. [3] H. Bushcek and A. J. Calise. Uncertainty modeling and fixed-order controller design for a hypersonic vehicle model. AIAA Journal of Guidance, Control, and Dynamics, Vol. 20. No.1, pp. 42-48,1997.5 [4] Q. Wang and R. F. Stengel, Robust nonlinear control of a hypersonic aircraft. Journal of Guidance, Control and Dynamics, Vol. 23, No.4, pp. 577-584,2000. [5] Y. Shtessel and J. McDuffie. Sliding mode control of the X-33 vehicle in launch and re-entry modes. in Proc. AIAA Guidance, Navigation, and Control Conference and Exhibit, 1998, pp. 1352-1362. [6] Austin, K.J. and Jacobs, P.A. (2001) 'Application of Genetic Algorithms to Hypersonic Flight control', IFSAWorld Congress and 20th NAFIPS International Conference, Vancouver, BC, Canada, July, pp.2428 {2433. [7] Xu, H., Mirmirani, M., and Ioannou, P., "Adaptive Sliding Mode Control Design for a Hypersonic Flight Vehicle," Journal of Guidance, Control, and Dynamics, Vol. 27, No.5, 2004, pp. 829-38. [8] Tournes, C., Landrum, D. B., Shtessel, Y., and Hawk, C. W., "Ramjet- Powered Reusable Launch Vehicle Control by Sliding Modes,"Journal of Guidance, Control, and Dynamics, Vol. 21, No.3, 1998,pp. 409-15. 4774 [9] Parker, J. T., Serrani, A., Yurkovich, S., Bolender, M. A., and Doman, D. B., "Control-oriented Modeling of an Airbreathing Hypersonic Vehicle," Journal of Guidance, Control, and Dynamics, Vol. 30, No.3, 2007, pp. 856-869. doi: 10.2514/ 1.27830 [10]Fiorentini, L., Serrani, A., Bolender, M., and Doman, D., "Nonlinear Robust/Adaptive Controller Design for an Airbreathing Hypersonic Vehicle Model," AIAA Paper 2007-6329,2007. [l1]Ge, S. S., and Wang, C. Adaptive NN control of uncertain nonlinear pure- feedback systems. Automatica, 2002, 38 (4):671-682 [12]Michael A. Bolender ,David B. Doman ,A Non-Linear Model for the Longitudinal Dynamics of a Hypersonic Air-breathing Vehicle,AIAA Guidance, Navigation, and Control Conference and Exhibit 15-18 August 2005, San Francisco, California' AIAA 2005-6255 [13]D. Swaroop, J. K. Hedrick, P. P. Yip and J. C. Gerdes, "Dynamic surface control for a class of nonlinear systems," IEEE Trans. Autom. Contr. Vol. 45 no. 10, pp. 1893-1899,2000. [l4]P. P. Yip and 1. K. Hedrick, "Adaptive dynamic surface control: asimplified algorithm for adaptive backstepping control of nonlinear systems," Int. J. Contr. Vol. 71, no. 5, pp. 959-979, 1998.