NONLINEAR FLIGHT CONTROL USING BACKSTEPPING TECHNIQUE

Acacio Morales1 , Pedro Paglione2 , Francisco Vargas3

Abstract: A nonlinear ight control system is proposed using backstepping. It is implemented a controller with an internal loop controls involving the angular rates of the aircraft and an external loop which includes angle of attack, sideslip angle and bank angle. Also, it is implemented a separated controller of velocity. Finally, nonlinear simulation results for a conceptual model of a medium size jet are presented to demonstrate the effectiveness of the proposed control law. Keywords: Nonlinear Control, Backstepping, Dynamical Systems. 1. INTRODUCTION Backstepping control design constitutes an alternative to feedback linearization. Using backstepping, system nonlinearities do not have to be canceled in the control law. If a nonlinearity acts stabilizing, and thus in a sense is useful, it may be retained in the closed loop system. This leads to robustness to model errors and less control effort may be needed to control the system [5]. A weakness of Backstepping as well as feedback linearization is that they lack support for dealing with actuator redundancy. The resulting control laws specify which total control effort to produce, but not how to produce it. For performance reasons, and also for safety reasons, modern aircraft are typically over-actuated in the sense that there are several combinations of control surface deections that will give the same aircraft response [4]. In this paper we introduce an alternative for control separation, a backstepping controller is used to achieve global stability with an internal loop controls involving the angular rates of the aircraft and an external loop which includes angle of attack, sideslip angle and bank angle without the two-timescale assumption to separate the slow and fast dynamics, then a separated control tracking of velocity using a rst order backstepping is implemented. In the section 2 is presented the aircraft model equation. This aircraft represents a medium size jet like EMBRAER

190/195 and Boeing 737-200/300 with augmented exibility. That model is selected because will be applied, in a posterior study, a nonlinear exible modes control to suppress the vibrational modes and lead the system to a ridig-body approximation, in fact, this preliminary study let us ensure that this nonlinear control theory will control the aircraft rigid ight dynamics to achieve global stability, also, to evaluate addition of looping controller using rst order backstepping. In the section 3, two Backstepping control design procedure are shown tracking four state variables. Finally, in the section 4 a numerical simulation is done to demonstrate a effectiveness of these control laws in two ight conditions. 2. AIRCRAFT MODEL The model of conceptual aircraft used in this paper is taken from da Silva et al. (2010a) [1]. It is based in a Doublet Lattice aerodynamic technique presented in Guimaraes Neto (2008) [3], which considers unsteady aerodynamics due to the structural vibration. However, in this study the structural vibration will not be considered, therefore the aircraft is consider as a rigid body system and equations of the amplitude exibility modes are not taken into account. The general nonlinear equations of motion over a at Earth are: V D T cos cos + g(sin cos sin + m m cos cos sin cos cos cos sin ) (1) L = q p cos tan r sin tan + mV cos g (cos cos cos + sin sin ) (2) V cos Y T = p sin r cos + sin cos + mV mV g (sin cos + sin sin cos cos sin V cos cos sin2 cos cos sin sin cos ) (3) = = I1 qr + I2 pq + I3 L + I4 N (4)

q r

= I6 r2 + I5 pr I6 p2 + I7 (M + MT ) = I2 qr + I8 pq + I4 L + I5 N

(5) (6)

The variables are respectively: inertial aircraft speed, angle of attack and sideslip angle V , , ; roll pitch and yaw rates p, q, r; lift drag and side aerodynamic forces L, D, Y ; thrust force T ; aerodynamic moments around x, y and z axes L, M, N and thrust pitching moment MT . The inertia moments Ii = 1, 2....9 are dened in ref [6], and depends of the aircraft inertia moments Ixx , Iyy , Izz and Ixz . Moreover, the six rigid body kinematic equations are: x0 = ((sin sin + sin cos cos ) cos sin + cos cos cos cos + (sin cos sin sin cos ) sin )V y0 = ((sin sin + sin cos cos ) sin sin + cos cos sin cos + (sin cos sin cos sin ) cos )V H = ((sin sin + sin cos cos ) cos cos cos sin )V = p + (q sin + r cos ) tan = q cos r sin = r cos sec + q sin sec (9) (10) (11) (12) (8) (7)

control inputs for the slow states , and intermediately. However, this methodology considers the transient responses of the fast states and, therefore, does not require the timescale separation assumption. First, it is necessary replace the aerodynamics forces and moments into the state equations 1 to 12. For example, force and moment around x are given by: c 1 2 V Sw (CL0 + CL a + 2 a 2V CLq q + el (CLei + CLeo ) + ail (CLasi + CLaso )) L = bw 1 2 Va Sw bw (Cl a + (Clp p + 2 2V Clr r + Cl ) + ail (Claai + Claao ) + rd (Clrl ) + Clru )) (16) (15)

L =

Where Sw , bw , V are wing area, wing span and equilibrium velocity (Used to calculate the stability derivatives) respectively. In 15 and 16 is used the aerodynamics results of angle of attack (a ), sideslip angle (a ) and Velocity (Va ), which are calculated using the following expression: = ui ug = vi v g = wi wg = = = u2 a arctan
2 + va wa

ua va wa Va a a

(17) (18) (19) +

Where H, x0 , y0 are altitude and inertial coordinates and , , and are pitch, bank and heading angles respectively. The aircraft possesses 8 aerodynamic controls plus the thrust control. This denes the control vector: u = [, ei , eo , asi , aso , aai , aao , rl , ru ] (13)

(20) (21) (22)

These controls, in the order of appearance: throttle setting, inboard elevator deection, outboard elevator deection, inboard aileron symmetric deection, outboard aileron symmetric deection, inboard aileron anti-symmetric deection, outboard aileron anti-symmetric deection, lower rudder deection, upper rudder deection. Aileron antisymmetric inboard and outboard are not commonly used to rigid body aircraft control, therefore, to simplify the model the inboard and outboard surfaces is coupled and antisymmetric aileron is neglected. Hence, the control vector can be reduced as: u = [, el , ail , rd ] (14) 3. CONTROLLER DESIGN The main objective of the controller is to track slow states V , , and . But the problem is divided in two parts. The rst one is the control of , and , this problem can be viewed like a two-scaletime approach because the fast states p, q, r are used as control inputs as is proposed in Lee et al. (2001) [6]. The second part is the control of V , accordingly to achieve this objective is proposed a rst order backstepping using the throttle setting as control inputs. 3.1. Controller of , and The Backstepping procedure to be applied can be viewed as two-timescale approach because the fast states are used as

ua arcsin va /Va

Where ui , vi , wi and ug , vg , wg are the vectorial components of inertial and gust velocity around of x, y and z axes respectively. Substituting all forces and moments into the state equations and rearranged to separate the slow T states variables x1 = [, , ] , the fast states variables T T x2 = [p, q, r] , the control variables u = [el , ail , rd ] T and x3 = [, ] yield: x1 x2 x3 = = = b1 (, , , ) + b2 (, , , )x2 + b3 (, )u c1 (, , p, q, r) + c2 (, )x2 + c3 (, )u d1 (, ) (24) (25) (23)

The state variables of altitude and inertial coordinates are not considered to this control application and the state variable V and control variable are used in the next controller design. The following are the most important assumptions used in the analysis process design: T Assumption 1: The desired trajectory xd = d , d , d 1 d d d are bounded as x1 , x1 , x1 cd where cd R is a known constant and denotes the 2-norm of a vector or a matrix.

Assumption 2: The magnitude of is bounded as || m < /2. Assumption 3: The control surface deection has not effect on the aerodynamic force component b3 (, ) = 0. Now we introduce the rst virtual control variable and its derivative: z1 z1 = = x1 xd 1 b1 + b2 x2 xd 1 (26) (27)

Now, we introduce a Lyapunov candidate function and its derivative: 1 2 z 2 V = zV zV =

Vl Vl

(39) (40)

Therefore, substituting 1, 37 and 38 into 40 we found the control input equation to make the result negative denite: D m (kv zV + + Vd g(sin cos cos cos m + sin + cos cos sin cos cos ) (41)

Considering the following Lyapunov candidate function: Td V1 V1 = 1 2 z 2 1 = z1 z1 = z1 (b1 + b2 x2 xd ) 1 (28) =

(29)

Furthermore, to calculate the input control is used the thrust expression from da Silva et al. [1] given by: = T i n Tmax (42)

Hence, making the Lyapunov derivative along of trajectory negative denite: xd = b1 k1 z1 b1 + xd 1 2 1 The second virtual variable is dened as: z2 z2 = x2 xd 2 = c1 + c2 x2 + c3 u xd 2 (31) (32) (30)

Where Tmax is the maximum Thrust of the model, is density, i is level sea density and n is a atmosphere model constant. 4. NUMERICAL SIMULATION This section presents numerical simulation results for each proposed controller to demonstrate the performance and effectiveness of this control law. A conceptual aircraft model of a medium size jet from [1] is used in this paper considering augmented exibility (conf2) and its controllability analysis shown in da Silva et al. (2010b) [2]. It is assumed that the aircraft is in a steady state level ight of V = 224.6m/s, H = 10000m. The controller structure is shown in gure 1. In order to obtain differential commands satisfying Assumption 1 in both controllers a third-order linear command lter is used and optimized to each command variable. The controller design constant chosen are: k1 = 9, k2 = 3 and kv = 1.15.
V Vd X1d + z1 zv
kv
Velocity Controller C

Now the Lyapunov candidate function is given by: V2 V2 = =

(33) (34)

Substituting the equations 26, 27, 31, 32 and 30 into 34 and in order to make the result negative denite we found the control input equation: u = c1 (k2 z2 c1 c2 x2 + xd A) 2 3 xd 2 (35)

Where A is the partial derivatives of and the transpose derivative of equation 23 as function of states and z1 , respectively, to avoid involving states derivative when the control is implemented and because is applied Lyapunov in matrix form [7]. xd xd 2 2 x1 x3 (36) A = bT z1 2 x1 x3 3.2. Velocity Controller In order to control the velocity is necessary determinate a control variable, in this case we want to determinate a necessary thrust to track velocity because we know that it depends of throttle setting which is a control variable; again assumptions 1, with desired velocity, and 2 are considered. At rst, we establish the virtual control and its derivative as: zV zV = = V Vd V Vd (37) (38)

k1

Outer Controller

X2d

z2 X2

k2

Inner Controller

Aircraft Ai ft Model

X1

Figure 1 Control laws proposed structure

In the rst simulation command input doublet values are applied to the system, d = 10 , d = 0 , d = 15 in 1s and d = 10 , d = 0 , d = 10 in 15s approximately, the initial condition is = 1.3 , = 0 , = 0 . In this simulation, it is not used the velocity control because in this case is evaluate short period response. The response is shown in gure 2.

15 10
el

10 V (m/s) 0 () -10 -20 -30

225.5 225 224.5 224 0 10 20 Time (s) 30 0 10 20 Time (s) 30

0.8 0.6 0.4 0.2 0 0 10 20 Time (s) 30

()

5 0 -5

10 20 Time (s)

30

10 5 0 -5

() 0 10 20 Time (s) 30

()

0.2 0.15 () () 0.1 0.05 0 -0.05 0 10 20 Time (s) 30

15 10 5

el

ail

-5

0 -5 -10 0 10 20 Time (s) 30 5 0 () -5 -10

10 20 Time (s)

30

1 ()
ail

0 -1 -2

20 15 () () 10 5 0 -5 0 10 20 Time (s) 30

15 10 5 0 () -5 -10 0 10 20 Time (s) 30

10 20 Time (s)

30 5 ()

10 20 Time (s)

30

10 5 0

rd

rd

-5 -5 0 10 20 Time (s) 30

10 20 Time (s)

30

Figure 2 Aircraft time response with a reference input. Blue () model response, green () reference input.

Figure 3 Aircraft time response into wind gust. Blue () with controller, green () without controller.

Another simulation is done, the aircraft is reached by a deterministic wind gust, controllers must maintain the system in the initial conditions, velocity controller is used. The result is shown in gure 3. 5. CONCLUSIONS In this paper nonlinear backstepping technique was used to design a control law to be applied in a nonlinear aircraft model with saturations and augmented exibility. A controller for a six-degree-of-freedom nonlinear was proposed to track , and using angular rates as intermediate, thus, it is possible control a slow dynamics using the fast dynamics. With this controller applied the error exponentially converges, then system can be globally stabilized or reach a new equilibrium state. A second backstepping controller was proposed to track V , in this case was controlled a slow variable using a separate control input, this system result not be very effective in short period but can be used in case of phugoid period or to remain the velocity when the aircraft is under wind gust condition. The use of backstepping for ight control design seems promising despite of a disadvantage is that the design complexity grows with the system order which makes it difcult to include, e.g., complex actuator dynamics, however, the results obtained shown that the technique is quite effective. ACKNOWLEDGMENT We thank those who supported us in any respect during the completion of the project.

REFERENCES [1] A. L. da Silva, P. Paglione, and Yoneyama T. Conceptual exible aircraft model for modeling, analysis and control studies. Submitted for presentation in AIAA., 2010. [2] A. L. da Silva, P. Paglione, and Yoneyama T. Controllability analysis and stabilization of exible aircrafts. XVIII Congresso Brasileiro de Automtica, 2010. [3] A.B. Guimaraes Neto. Dinmica e controle de aeronaves exveis com modelagem aerodinmica pelo mtodo doublet lattice. Masters thesis, Technological Institute of Aeronautics (ITA), 2008. [4] Ola Harkegard. Backstepping and Control Allocation with Applications to Flight Control. PhD thesis, Linkoping Studies in Science and Technology, 2003. [5] M. Krstic, I. Kanellakopoulos, and P. Kokotovic. Nonlinear and Adaptive Control Design. 1995. [6] Taeyoung Lee and Youdam kim. Nonlinear adaptive ight control using backstepping and neural networks controllers. Journal of Guindance, Control and Dynamics, 24, 2001. [7] Van Orrt Eduard Richard. Adaptive Backstepping Control and Safety Analysis for Modern Aircraft. PhD thesis, Technische Universiteit Delft, 2011.