Journal of Space Technology, Vol 7, No 1, July 2017

LQR and LMI Based Optimal Control Design for

Aircraft
Abdur Rasheed

controlling longitudinal motion and pitch while for lateral

Abstract—The maneuvering of an aircraft requires design control, aileron and rudder perform their role [4].
and implementation of advance control techniques for both The aircraft being highly coupled and non-linear dynamical
longitudinal and lateral motion. The main objective of this system is difficult to modeled accurately. Therefore, certain
paper is to design an optimal controller for aircraft dynamics assumptions are made to make derivation of model easy and
in terms of quadratic performance index. The problem retaining the model accuracy within desired limit. Aircraft is
formulation is carried out using different control techniques assumed to be a rigid body with constant mass and also some
like linear quadratic regulator (LQR) and linear matrix deviations exist in motion from equilibrium level flight [5].
inequalities (LMI). The controller design process begins with The mathematical modeling of aircraft can be carried out
obtaining appropriate mathematical model for aircraft using first principle approach or using system identification
dynamics. The model is obtained using first principle technique. The first principle method is a theoratical method
approach and linearized using small perturbation theory. Open uses Newton-Euler equations for description of dynamic
loop analysis is carried out for model using MATLAB and characteristics of aircraft. This paper also considers this
results are obtained in step responses. The model response is method for derivation of longitudinal motion equations which
compared with LQR and LMI based results.The simulation are linearized using small perturbation theory.
results reveal the effectiveness of these control techniques The controller design for aircraft dynamics which are
applied and ensure stability of aircraft dynamics. Longitudinal coupled and highly non-linear is a difficult task and it
model of aircraft is used for analysis and aircraft used as becomes further difficult due to its sensitivity to disturbance.
reference is Cessna 310. Research interest for control design techniques preferred to
achieve desired performance without causing increase in
Keywords— Aircraft, LMI, LQR, Longitudinal, Cessna 310. complexity. Modern control methods exist for overcoming all
short comings exist in flight control of aircraft but
I. INTRODUCTION proportional integral derivative is still in use because of its
ease in implementation and reduced cost [6]. For flight control
The development and advancements of aircraft of various system to maintain robustness along with maintaining desired
design is still a research area of growing interest because of its performance uses a robust control technique named ⁄
civil and military applications. The challenges that any aircraft [7,8]. The main idea of using this method is to maintain
goes through during its mission assigned not only depend boundary limit between performance of aircraft and
upon the aerodynamic design but also on system design. robustness. State output feedback (SOF) is another method for
Aircraft of light weight low cost with efficient control autopilot design of aircraft dynamics and its significance of
techniques and strategies are of great interest. Effective this method is that it can work on signal information to be
control methods are applied during designing which can be controlled. The SOF technique calculate static gain in order to
helpful in case of failure of control surface, actuator etc and meet desired requirement for system in closed loop [9].
also provide with efficient performance during its flight [1]. Dynamic inversion is among other methods which is applied
in control community and is mainly energized due to
Aircraft is a highly non-linear and complex system. Free aerospace control significance [10- 13]. It helps in retaining
flight motion for any aircraft is very complicated task [2]. non-linear characteristics along with flexibility and its
simplicity. LQR is another controller design approach used to
The aircraft motion basically consists of three translational achieve stability and control of aircraft motion. It can maintain
and three rotational motion along x-, y- and z-axes. The three trim state of aircraft if uncertainties are generated due to loss
translation motion includes horizontal motion along x-axes, of control surface. LQR method calculate feedback gain which
vertical motion along z-axes and transverse motion along y- stabilizes the motion and overcome disturbance uncertainties
axes while three rotational motion includes roll, pitch and yaw [14-15].
motion. The main control surfaces designed for control of The organization of paper is carried out as: The first part of
different types of motion are elevator, rudder and aileron. paper discusses modeling of aircraft longitudinal motion using
Aileron located on wing controls pitch motion, rudder on first principle approach with linearization of non-linear
vertical tail affect yaw motion and elevator is responsible for differential equations using small perturbation theory and
roll motion [3]. The aircraft motion can be mainly categorized obtaining state space form. The open loop analysis of model is
into longitudinal and lateral motion resulting in two main carried out and simulation results are obtained for step
types of control i.e. longitudinal control and lateral control. In responses. The next part describes control technique named
case of longitudinal control, elevator plays its role in linear quadratic regulator (LQR). The LQR calculated gain

CECOS University, Peshawar, Pakistan.

Email: [email protected] 97
 LQR and LMI Based Optimal Control Design for Aircraft
matrix stabilize the system dynamics and results obtained are
compared with that of open loop. The second part of the third ( )⁄
∑ (1)
section deals with another control method named linear matrix
inversion (LMI) which is a static feedback controller design, ∑ ⁄ (2)
stabilizing the system as evident from simulation results. LQR
and LMI simulation results are compared and conclusion is Disturbance act upon aircraft dynamics and its forces and
drawn in last section. moment equations get modified to following form:

II. AIRCRAFT DYNAMICS ∑ ∑ ∑ (3)

This section describes mathematical modeling of aircraft ∑ ∑ ∑ (4)

which makes easier to understand characteristics of aircraft.
The different axes of aircraft are shown in figure 1 describing Where ∑ and ∑ represent sum of steady state forces
roll, pitch and yaw motion. The geometric and flight data and moments while sum of disturbance forces and moments
regarding aircraft is shown in table I. are described by ∑ and ∑ respectively. The main
The different assumptions made for derivation of reason for disturbance are control surfaces deflection or
mathematical model of aircraft are; it is assumed to be a rigid atmospheric disturbances.
body with constant mass and small disturbances exist in its The equations of motion describing aircraft dynamics are as
motion. The aircraft equations of motion are derived using given as follows;
Newton-Euler equations and these equations are as follow [16-
17]; ( ̇ )
[ ] [ ( ̇ ) ] (5)
( ̇ )
̇ ( ̇ ) ( )
[ ] [ ̇ ( ) ( )] ( )
̇ ̇ ( )

Where and are components of translational velocity, p,

q and r are components of rotational velocity, m is mass of
aircraft and , and are components of gravity along
, and axes. Similarly, , and are rolling,
pitching and yawing moments along x-, y- and z-axes. The
forces, moments, linear velocity and angular velocity
components are shown in figure 2.

Fig. 1. Aircraft axis

Table I
Geometric and flight data
Mass (lbs) 4600
Mean aerodynamic chord (ft) 4.79
Wing Surface ( ) 175
Wing Span (ft) 36.9
Altitude (ft) 8000
Mach Number 0.288
Air Speed ( ⁄ ) 312.5
Dynamic Pressure ( ⁄ ) 91.2
Location of CG % of MAC 0.33
Moment of inertia x-axes ( 8,884
)
Moment of inertia y-axes ( 1,935
)
Fig. 2. Forces, Moments and velocities components
Moment of inertia z-axes ( 11,001
)

98
 LQR and LMI Based Optimal Control Design for Aircraft
Euler angles describes the aircraft motion in inertial space
and equations for ̇ , ̇ and ̇

̇ (7)
̇ (8)
̇ (9)

The linearization of non-linear equations is carried out using

small perturbation theory and linearized equation are
converted to state space representation. The generalized state
model is given as;

̇ (10)
(11)
Fig. 4. Response for pitch rate
Where A, B, C and D describes system matrix, control
matrix, output matrix and feed forward matrix. In our case, the
state vector for longitudinal model is given by
[ ] , Where u is longitudinal component of true
airspeed, is the angle of attack, q is pitch rate and is the
pitch angle. The elevator control surface is used as input
which affects the longitudinal motion. The numeric values
obtained for the matrices A and B of the longitudinal model of
the aircraft are:

A [ ]

B [ ] Fig. 5. Response for pitch angle

The simulation results in fig. 3-5 shows quite unsatisfactory

behavior for the mentioned state. The angle of attack, pitch
The eigen values obtained for the longitudinal model are: - rate and pitch angle should obtain zero as their steady-state
9.93e+000, -7.44e-001, -4.79+1.84e-002i and -4.79+1.84e- value according to trim conditions.
002i. The open loop response obtained for the model shows Similarly the frequency response of the open loop longitudinal
unsatisfactory performance and these simulations are shown in model can be studied from fig. 6.
fig. 3-5.

Fig. 6. Longitudinal model frequency response

Fig. 3. Response for angle of attack

99
 LQR and LMI Based Optimal Control Design for Aircraft
These unwanted characteristics should be eliminated anyhow cost function is sum of terminal cost and integral along way is
and it became necessary to design controller(s) so that desired given by;
results can be achieved. These different control strategies
designed and implemented are discussed in the next section. ( ( ) ) ∫ ( ( ) ( ) ) (14)

Where terminal cost is described by ( ( ) ) and

III. FLIGHT CONTROLLERS DESIGN
( ( ) ( ) ) is non-negative. Also ( ( ) ) for
linear quadratic regulator and L is given as;
The simulation results obtained for the longitudinal model of
the aircraft show diverging behavior. These states should be
stable and required to obtain their desired steady state value ( ) ( ) (15)
within the required time. In order to achieve desired
performance the different control techniques implemented The feedback control law obtained on the basis of linear
using Matlab/Simulink are Linear Quadratic Regulator (LQR) model assumptions is given as;
and Linear Matrix Inequality (LMI). These control methods
are described in the following sections. ( ) ( ) (16)

A. Linear Quadratic Regulator (LQR) The value of gain matrix K obtained for longitudinal model
is given as;
Optimal control design approach for multi-input multi-output
(MIMO) dynamical systems is one of the important control [ ]
method which work effectively for systems. LQR is a control
strategy which when design properly provide with desired The LQR controller stabilize the plant model very effectively.
performance of a system which can be stated in terms of After introducing K into the model dynamics the simulation
performance index. It provides optimal approach to calculate results obtained for longitudinal model are shown in fig. 8-10.
feedback gain and for that all states of systems should be
available for controller. The block diagram for LQR is shown
in figure 7.

Fig.7. LQR Block diagram

Fig. 8. LQR response for angle of attack

The feedback control law implemented for the state space

form of the aircraft dynamics in (10) is given as [18-19];

( ) (12)

After obtaining value for gain matrix K, the longitudinal

model dynamics modifies to following form:

̇ ( ) (13)

is input to external closed loop system and

representing desired state vector. The modified system matrix
for plant dynamics after K calculation is A-BK. The
performance index J which need to be minimized and gives
Fig. 9. LQR response for pitch rate

100
 LQR and LMI Based Optimal Control Design for Aircraft

The system performance shows significant improvement

using LMI based dynamic model. The effect of disturbance is
accounted effectively. The performance shows high degree
efficiency of LMI based control approach. The simulation
results obtained are shown in fig. 11-13.

Fig. 10. LQR response for pitch angle

The simulation results obtained in fig. 8-10 using LQR

control technique shows significant performance in improving
the response and characteristics of the longitudinal model. The
angle of attack response in fig. 8 has zero steady state value
according to requirement without any delay of time. Similarly,
for pitch rate and pitch angle, their desired steady state value Fig. 11. LMI simulation for angle of attack
is achieved efficiently. There is slight variation from their trim
value at the beginning but after short time of few second, the
desired final value is attained.

B. Linear Matrix Inequality (LMI)

For an affine function with decision vector

[ ] and with real symmetric matrices
, the matrix inequality is given as[20-22];

( ) + >0 (17)

is known as Linear Matrix Inequality (LMI).

The disturbance showing impact on longitudinal motion is
given by [ ] hence affecting horizontal velocity,
angle of attack, pitch rate and pitch angle. The main objective Fig. 12. LMI simulation for pitch rate
is to design control law of form: which guarantees
the system stability and its performance index J is given by;

∫ ( ) ( ) ( ) (18)

The Lyapunov stability matrix P obtained is given as;

P [ ]

The gain matrix K calculated for state feedback is given as;

[ ] Fig. 13. LMI simulation for pitch angle

101
 LQR and LMI Based Optimal Control Design for Aircraft

The simulation results in fig. 11-13, obtained for the

longitudinal model of the aircraft using LMI control strategy
have significantly improved its performance. The desired
steady-state results are achieved effectively and system
behavior modified efficiently.

IV. LQR and LMI PERFORMANCE COMPARISON

The Linear Quadratic Regulator (LQR) and Linear Matrix

Inequality (LMI) control strategies designed for the
longitudinal model of the aircraft have brought significant
improvement in the performance of the model obtained. The
unwanted behavior of the open loop model is eliminated
efficiently and effectively. The performance of both LQR and
LMI are compared in terms of their time domain responses to Fig. 16. LQR-LMI comparison for pitch angle
reach their desired steady state values. Their comparison is
shown in fig. 14-16. From fig. 14-16, it is evident that the LMI control strategy
designed effectively and efficiently attains the desired trim
value for each state of the model. The LQR controller
performance for pitch rate and pitch angle undergoes through
slight variation from the steady-state value and attain it after a
short interval of time. Overall, the performance of the LMI is
much effective than the LQR.

V. CONCLUSION

This paper has considered feedback stabilization for

longitudinal motion of aircraft through LQR and LMI
approach. The 6 degree-of-freedom model of an aircraft is
derived using first principle approach and these non-linear
equations are linearized using small perturbation theory and
converted to state space model. The stability analysis is
carried out for open loop system and simulation results are
obtained for step responses. The control design for technique
developed are linear quadratic regulator and linear matrix
inversion. The solution of the problem with both LQR and
Fig. 14. LQR-LMI comparison for angle of attack
LMI approach is demonstrated and results obtained shows that
the performance and handling qualities are satisfactory with
the reference model over wide range of nonlinear flight
regime.

REFERENCES
[1] M.V. Cook,(2007). “Flight Dynamics Principles: A Linear
SystemsApproach to Aircraft Stability and Control”,Elsevier Ltd..
[2] R.C.Nelson,(1998). “Flight Stability and Automatic Control”,McGraw
Hill, Second Edition.
[3] M. Drela, “Flight Vehicle Aerodynamics”, MIT Press Ltd, USA, 2014.
[4] N. Wahid, N. Hassan and M.F. Rahmat,“Application of Intelligent
Controller in Feed-back Control Loop for Aircraft Pitch Control”,
Australian J. of Basic and Applied Sci.-2011.
[5] L. Sonneveldt, E.R. Ort, Q.P. Chu and J.A. Mulder, “Nonlinear Adaptive
Flight Control Law Design and Handling Qualities Evaluation”,IEEE
Conf. on Decision and Control, China, Dec. 16-18, 2009.
[6] B. Song, J. K. Mills, Y. Liu, and C. Fan, “Nonlinear Dynamic Modeling
and Control of a Small-Scale Helicopter”,Int. J. Control Autom. Syst.,
Fig. 15. LQR-LMI comparison for pitch rate vol. 8, no3, pp. 534-543, 2010.
[7] R. Lozano, “Unmanned Aerial Vehicles Embedded Control”, Wiley-
ISTE, 2010.

102
 LQR and LMI Based Optimal Control Design for Aircraft
[8] E. Promtun, “Sliding Mode Control of F-16 Longitudinal Dynamics”,
MS. Thesis, SanDiego State University, San Diego, USA, 2007.
[9] O. Rehman, I. Petersen, and B. Fidan, “Robust Nonlinear Control
Designof a Nonlinear Uncertain Systems with Input Coupling and its
Application to hypersonic flight vehicles”, Multi-Conference on
Systems and Control, Yokohama, Japan, 2010.
[10] N. Kishor, R. P. Saini and P. Singh, “LQG/LTR Controller for Speed
Governing of Hydro-Turbine”,12th IEEE Electrotechnical Conference in
Melecon,vol.3,pp 1125-1128, May 2004.
[11] M. Lv, Y. Hu and P. Liu, “Attitude Control for Unmanned Helicopter
using H-infinity Loop Shaping Method”, International Conference on
Mechatronics Science, Electrical Engineering and Computer, Jilin,
China, Aug. 19-22, 2011, pp.1746- 1749.
[12] Q. Zheng, L. Dong and D. H. Lee, “Active Disturbance Rejection Control
for MEMS Gyroscopes”, IEEE Trans. on Control Systems
Technology, Nov, 2009, VOL. 17, No. 6: 1432-1438.
[13] Zhiqiang Gao. “Active Disturbance Rejection Control: A Paradigm Shift
in Feedback Control System Design”, American Control Conference,
Minneapolis, Minnesota, USA, June,2006.
[14] M. M. Almeida and G. V. Raffo, “Nonlinear Control of a Tiltrotor UAV
for Load Transportation,” in 11th IFAC Symposium on Robot Control,
August 2015, pp. 234–239.
[15] J.K. Shiau and D.M. Ma, “An Autopilot Design for the Longitudinal
Dynamics of a Low Speed Experimental UAV using Two Time Scale
Cascade Decomposition”,Transaction of the Canadian Society for
Mechanical Engineering, Vol 33, No 3, 2009.
[16] R. F. Stengel, “Flight Dynamics” Princeton University Press, 2004.
[17] J. Anderson,“Fundamentals of Aerodynamics”, 5th Ed. Mcgraw Hill
Series in Aeronautical and Aerospace Engineering, 2010.
[18] N. Wahid, M.F. Rahmat, “Pitch Control System Using LQR and Fuzzy
Logic Controller”, Industrial Electronics & Applications (ISIEA), IEEE
Symposium on Industrial Electronics Malaysia,2010, pp.389-394.
[19] N. Wahid, M.F. Rahmat, K. Jusoff, “Comparative Assessment Using
LQR and Fuzzy Logic Controller for a Pitch Control System”. European
Journal of Scientific Research, Vol 42, 2010, No 2, pp.184-194.
[20] S. Boyd and L. Vandenberghe, “Convex Optimization”, Cambridge
University Press, Cambridge, U.K.
[21] Cai, G. B., Hu, C. H., and Duan, G. R. Efficient LMI-Based Quadratic
Stability and Stabilization of Parameter-Dependent Interval Systems
with Applications”, International Journal of Innovative Computing
Information and Control,8(3A):1943–1954, 2012.
[22] A. Casavola, D. Famularo, and G. Franze, “Robust Fault Detection of
Uncertain Linear Systems via quasi-LMIs”, Automatica, 44(1):289–295,
2008.

103