Recent advances on Mathematical Models for Engineering Science

Design of a Lateral Motion Controller for a

Small Unmanned Aerial Vehicle (SUAV)
Ahmed Elsayed Ahmed, Electrical Engineering Dept., Shoubra Faculty of Engineering, Benha
University, Egypt (Telephone: +201007124097), E-mail: eng_medoelbanna @yahoo.com.

A. N. Ouda, B. Military technical college, Cairo, Egypt, E-mail:[email protected]

Ashraf Hafez, Electrical Engineering Dept., Shoubra Faculty of Engineering, Benha University,
Egypt. E-mail: [email protected].

Hossam Eldin Hussein Ahmed, Communication Engineering Dept., Faculty of Electronic Engineering,
Menoufia University, Menouf, Egypt (fax: +20483660716), E-mail: [email protected].

Hala Mohamed Abd-Elkader, Electrical Engineering Dept., Shoubra Faculty of Engineering, Benha
University, Egypt, E-mail: [email protected].

Abstractthe unmanned aircraft systems (UAS) are playing increasingly prominent roles in defense programs and strategies around the
world. This paper presents a design of lateral autopilot of a Small Unmanned Aerial Vehicle (SUAV). The designed autopilot is applied to an
Ultrastick-25e fixed wing UAV depending on lateral linear model and analytic linear model of a coordinated turn derivation with trimmed
values of straight and leveling scenario. The lateral motion controller design is started with the design of most inner loop (roll rate feedback) of
the lateral system, then roll tracker design with a Proportional Integral (PI)- controller. The guidance and control system is related with the
design of heading direction controller with P-controller. Yaw damper is designed with washout filter to maintain a zero sideslip angle. The
performance of two classic controller approaches for the design of autopilot are compared and evaluated for both linear and non-linear models.
The results show a good performance in both disturbance rejection and robustness against sensors noise.

Keywords autopilot, inner loop, outer loop, roll tracker, yaw damper, heading direction controller.
I. INTRODUCTION
input controls ( ( ), ( )). In general
U AV Automatic Flight Control System (AFCS) presents a
very comprehensive treatment of UAV control and
related technologies. The inherently unstable nature of typical
terms, an autopilot is a system used to guide an aircraft
without the assistance of a pilot. For manned aircraft, the
open loop UAV configurations necessitates a rigorous autopilot can be as simple as a single axis wing-leveling
approach to the analysis and design of UAV control autopilot. a full complete complicated autopilot as a full flight
technologies, as well as a thorough understanding of stability control system that controls position (altitude, latitude,
issues. longitude) and attitude (roll, pitch, yaw) during the various
For any dynamic system, there may be requirements on phases of flight (e.g., take-off, ascent, level flight, descent,
certain frequency (to ensure the dynamics are fast) and landing, and loiter).
damping (to ensure that the oscillations die out quickly)
specifications on the pole locations, and minimizing steady For SUAVs, the autopilot is in complete control of the
tracking error to commanded input. The typical solution to aircraft during all phases of flight. From beginning, the design
satisfy these requirements is to use linear feedback control to of autopilot is separated into two separate designs;
modify the pole locations and loop gains [1]. longitudinal and lateral motion controllers [3]. These separate
The complete state of the UAV comprises its position, autopilots are designed after modeling of ultrastick-25e fixed
airspeed (Va), attitudes (roll (), (), ()), angle-of- wing UAV [4], the longitudinal dynamics (forward speed,
attack (), sideslip angle (), and rotation (roll (p), pitch (q), pitching, and climbing/descending motions) and the lateral
and yaw (r)) rates. Position, airspeed, and attitudes are also dynamics (rolling, and yawing motions). This simplifies the
known as the navigation state [2].The main purpose of this development of an accurate autopilot simply.
paper is to design a lateral motion controller for SUAVs to SUAV control and stabilization is more difficult than larger
completely control lateral states (, , , , ) by the lateral one. This is due to several factors, including the low mass of

ISBN: 978-1-61804-341-2 102

 Recent advances on Mathematical Models for Engineering Science

the vehicle, lower Reynolds numbers, and light wing loading. pass filter action of the washout filter is to remove this steady
These factors make it more difficult to design a flight control state component. The output of the washout approximates a
system [5]. differentiated yaw rate which is suitable feedback for the
In order to fly an aircraft, low-level inner loops must dutch roll mode [8].
stabilize the airframe using available sensors and actuators, For large angle of attack another problem can be appeared
higher-level outer loops control will implement path following which is that the roll pole coupled with the spiral pole to form
(guidance), while the inner loop keeps the aircraft flying a complex pair, the solution of this is to deal with the system
(stabilization). Detailed design procedures of a flight control as multi input multi output system due to the coupling
system to adequately stabilize and control SUAV are utilized between the inputs and the outputs at high angle of attack, or
we can assume that the maximum angle of attack is 20
by designing autopilot inner loops, and outer loops.
degrees. The last assumption is good and suitable for laminar
The two derived linearized models (state space and
conditions because great values of angle of attack can cause
analytical linearized models) for Ultrastick-25e (thor) aircraft
turbulent so the stolen condition will be happened to fail the
were used in the design of autopilot [4]. The behavior of the flight of the aircraft [9].
aircraft due to the desired scenarios results were compared After that the controller of roll attitude (bank angle hold
between (the state space linearized model and the analytical controller) will be designed from the inner loop of roll rate by
linearized model) and (the nonlinear aircraft dynamics), the PI-controller.
results is too matched between all of the three. The last controller is the outer loop and used for UAV
The outer-loop was designed to achieve the tracking guidance; the heading direction of aircraft with P-controller
command requirements in (ground track angle). The inner only which satisfies the designed purpose. The whole lateral
loops are designed to track roll attitude reference signals motion controller block diagram as shown in Fig. 1
required for the outer loop. Several design goals were
introduced against the inner loop performance that the closed
loop rise time should be less than 1 second, and the overshoot
has to be smaller than 5% in some cases. Root locus technique
and PI-controller were tuned using the two linearized models
of the aircraft [6].
If the controller causes overshoot and degrades the
controller performance when coming out of saturation. In
order to prevent this, an anti-windup scheme is implemented Fig. 1: Lateral motion controller block diagram.
which checks if the actuator would saturate on the current time
step and does not perform the integration if this case is III. INNER LOOPS OF LATERAL MOTION CONTROLLER (ROLL
happened [7].Roll angle reference is constrained at 45. ATTITUDE TRACKER) DESIGN.
This paper is organized as follows; 1st section is the
Roll rate feedback is designed to increase the damping of
introduction, 2nd section investigate the design procedures of a
roll rate and is the most inner loop of the roll attitude (bank
lateral motion controller, 3rd section is the design of the inner
angle) hold controller. The analytical open loop linearized
loops of lateral motion controller, 4th section shows the inner
transfer function of roll rate from aileron is used to design the
loops design results, 5th section is the design of the heading
roll damper then bank angle tracker. After that the main test
direction controller, 6th section is the design of the yaw
results.
damper, 7th section shows the performance of the whole lateral
Roll rate (p) from aileron transfer function can be
motion controller in the scenario of rectangular motion
approximately simplified to a reduced linear transfer function
command to the aircraft, and 8th is the last section which is the
conclusion. (roll mode) [4, 10] considered as eqn. (1).
156 .5
() (+16.09) () (1)
II. LATERAL MOTION CONTROLLER DESIGN PROCEDURES The chosen pole will move from -16.09 to the left to be -
The lateral motion controller uses the rudder and ailerons to 24.8 with kd_p = - 0.055, the gain value is chosen to smooth the
keep the aircraft flying in a coordinated turn and following a response trajectory of roll tracker.
commanded turn rate (including a zero turn rate for straight After that, the next inner loop is the roll angle () tracker
flight), it consists of inner loops and outer loop to manage the by integrating the roll rate transfer function to extract the roll
navigation of the aircraft in the lateral motion and zero side attitude, and design the controller using PI- controller
slip angle control, the dynamics affected the lateral motion MATLAB structure as in Fig. 2.
begin with the body axis roll rate which is fed back to the The performance analysis of roll tracker with two
ailerons to modify the damping of the roll mode. controllers in linear and nonlinear models is showed in the
Yaw rate feedback is designed to modify the damping of the next section. The UAV Laboratory Web site provides detailed
dutch roll mode, but yaw rate feedback only is not sufficient
information on the airframe, avionics, and software
due to coupling between yaw and roll which results a steady
architecture [11]. The flight software is available as an open
state yaw rate component during turns, a simple solution is use
source software package available upon request, this open
a washout filter on the output of the yaw rate sensor. The high
source package has a classic PID controller for autopilot, and

ISBN: 978-1-61804-341-2 103

 Recent advances on Mathematical Models for Engineering Science

the proposed autopilot design modifies the performance of the designed controller parameters (kd_p = -0.055, kp_ = -0.89,
classic one. Again the classic controller parameters for roll and ki_ = -0.45).
tracker (kd_p = -0.07, kp_ = -0.52, and ki_ = -0.2) and the

1
-K-
s
ki_phi Int1
rad2deg
0.4363 -K-
|d/dt| 8.0 Hz P -156.5 1
-K- bias/scf/noise
<2.6 BW L-0.4363 s+16.09 s
phi
kp_phi p/delta_a phi_int
Aileron Servo

disturbance
kd White Noise

-.055

phi_ref the classic parameters the designed parameters

kd_p = -0.07 kd_p = -0.055
kp_phi = -0.52 kp_phi = -0.89
ki_phi = -0.2 ki_phi = -0.45

Fig. 2 MATLAB structure of bank angle hold controller.

When the input (desired output) in the roll attitude is five

IV. INNER LOOPS OF LATERAL MOTION CONTROLLER TEST degree doublet signal, the output responses with the two
RESULTS. controllers are shown in Fig. 3 where the designed controller
The tests executed are the following tests: graph is better than the classic assuring that the time domain
- Time domain analysis. characteristics in Table I.
- The effects of sensors noise in the response.
TABLE I
The effect of the sensors noise with sigma = 1.0*10-03 rad
TIME DOMAIN ANALYSIS OF ROLL TRACKER
/sec in the response of the roll attitude as shown in Fig. 4
The property Classic con. Designed con. 8
doublet response for roll tracker in the existence of noise

designed controller
tr [sec] 0.4194 0.2568 6 classical controller

ts [sec] 3.1875 3.2009 4

roll angle[deg/sec]

Max.o.s% 16.2839 7.2120 2

Undershoot% 0 0 0

Peak 1.1628 1.0721 -2

Peak time [sec] 1.1737 0.7733 -4

-6
- Doublet signal response.
-8
0 5 10 15 20 25 30
Doublet response comparison between the classical and designed controller fo roll tracker time[sec]
6
designed controller
classical controller
Fig. 4 The effect of noise in the roll tracker
4 - The effect of disturbance in the doublet response.
2 The ability of the disturbance rejection between two
roll angle[deg/sec]

controllers is shown in Fig. 5 where the designed controller

0
response to disturbance is high in the value, but it attenuated
-2 more speed than the classic controller.
-4

-6
0 2 4 6 8 10 12 14 16 18 20
time[sec]

Fig. 3 +5 degree doublet signal in roll response.

ISBN: 978-1-61804-341-2 104

 Recent advances on Mathematical Models for Engineering Science

12
+5 deg response for roll tracker in the existence of disturbance
By the previous tests, the roll tracker of the designed
designed controller
classical controller controller assures that it has a good performance and good
10
robustness and disturbance rejection; its closely matched with
8 the desired requirements. The next controller is the direction
roll angle[deg/sec]

heading controller, its considered as an outer loop of the roll

6
tracker and related with the guidance and control navigation
4 system of SUAV.
2
V. OUTER LOOP LATERAL MOTION CONTROLLER (HEADING
0
0 2 4 6 8 10 12
HOLD CONTROLLER) DESIGN.
time[sec]
From equation (2) [10], the heading rate is related with the
Fig. 5 The effect of disturbance in the roll tracker.
pitch rate q and yaw rate r and pitch attitude and roll attitude
- Multi-step of roll attitudes response. . From this equation we can get a new simplified linear
16
multi step response for roll tracker relation between the heading angle and the roll angle, this
14
designed controller relation is obtained from the coordinated turn assumption [12].
classical controller
During a coordinated turn, in the absence of wind or sideslip,
we have that = , and = .
12

10
roll angle[deg/sec]

= +
8
(2)
6
= n = = n (3)
4

Equation (3)can be rewritten as:
2

0
= + (tan )
-2

-4
0 5 10 15 20 25 30 35 40
= +
time[sec]

Fig. 6 Multi-steps response of roll tracker By expressing the equation in Laplace form
Multi steps response is used for commanding the aircraft
= () + () (4)
many sequences of heading angle as will be seen in the
rectangular motion, from Fig. 6 it can be shown that the Equation (4) is the last expression that we can use it to
response to the designed controller is faster than the classic control the heading from roll attitude; Fig. 7 illustrates the
one. MATLAB structure block diagram of the lateral autopilot.

1 1
0 -K-
s s
ki_psi Int1 ki_phi Int2
0.4363
-K-
|d/dt| 8.0 Hz P -156.5 1 9.81/17
1.25 -0.89 bias/scf/noise
BW L s
<2.6 s+16.09 s psi
-0.4363 rad2deg
kp_psi kp_phi roll Int psi int
Aileron Servo p/delta_a
White Noise

kd_p

-0.055

the classic parameters are the designed parameters are

kd_p = - 0.07 kd_p = - 0.055
kp_phi = - 0.52 kp_phi = - 0.89
ki_phi = - 0.2 ki_phi = - 0.45 disturbance
kp_psi = 1.2 kp_psi = 1.25
ki_psi = 0 ki_psi = 0
psi_ref

Fig. 7 MATLAB structure block diagram of the lateral autopilot.

classic and designed controllers is matched so much, but the
VI. LATERAL DIRECTION CONTROLLER TESTS RESULTS contribution here is the analytical linearized model and a
After designing the heading controller, the following tests controller with a new Simulink model and compares it with
were executed to check its performance as shown in the the Minnesota university test platform [11]. The first test is the
following figures (8, 9, and 10); the comparison between the time domain specs, then linear analysis, and the nonlinear

ISBN: 978-1-61804-341-2 105

 Recent advances on Mathematical Models for Engineering Science

analysis with a scenario of rectangular motion. 12

The effect of disturbance in the heading hold controller

10

- Time domain analysis. 8

heading angle [deg]

6
TABLE II
4
TIME DOMAIN ANALYSIS OF ROLL TRACKER
2

The property Classic con. Designed con. 0

tr [sec] 0.6308 0.5826 -2

ts [sec] 3.2707 0.9220 -4

Max. O.S. % 14.1546 1.3984 -6

0 2 4 6 8 10
time[sec]
12 14 16 18 20

Undershoot% 0 0
Fig. 10 The effect of disturbance in the heading hold
Peak 1.1415 1.0140 controller.
Peak time [sec] 1.4129 1.2066 The disturbance signal is a signal with amplitude 5 degree
From Table II the designed controller is better than the and pulse width of 0.2 sec applied after steady state of the
classic one in the rise time and settling time, and maximum response at a time from 8 to 10 Seconds. From Fig. 10, the
overshoot (this property is obvious in the step response test). two controllers are good in the rejection of disturbance.
- Doublet signal from standalone Simulink model. In the test platform linear simulation; the executed scenario
+5 degree Doublet response of the heading angle psi
is the rectangular motion, the command and response is shown
5
designed controller
in Fig. 11.
4 classical controller Rectangular motion with heading angle psi
400
3

2 350
heading angle [deg]

1 300

0
heading angle [deg]

250

-1
200
-2
150
-3

100
-4

-5 50
0 5 10 15 20 25 30 35 designed controller
time[sec] classical controller
0

Fig. 8 Doublet signal response of the heading hold -50

0 10 20 30 40 50 60 70 80
controller. time[sec]

Fig. 8 illustrates the matching between two controllers if Fig. 11 Rectangular motion heading command and its
the input is a doublet response. response in linear simulation model.
- The effect of sensor noises.
The noise test is executed under the noise at sigma= 1*10-3 VII. YAW DAMPER DESIGN
rad/s Yaw rate is fed back to the rudder to modify the dutch roll
The effect of noise in the heading angle
6 mode. The purpose of the stability augmentation of the yaw
rate feedback is to use the rudder to generate a yawing
4
moment that opposes any yaw rate that builds up from the
dutch roll mode. The resulting feedback is a type 0 yaw rate
heading angle [deg]

with zero command input [13].

0
We will use a washout filter on the output of the yaw rate
-2
sensor to remove the steady state components of the yaw rate
during turns
-4
Washout filter time constant is a compromise; too large
-6
value is undesirable since the yaw damper will then interfere
0 5 10 15 20 25 30 35 40 45 50
time[sec] with the entry into turns [1]. Too small value will reduce the
Fig.9 The effect of noise in the heading hold controller. achievable dutch roll dumping. Time constant () was chosen
- The effect of disturbance (the ability to disturbance to be 0.5 s.
The transfer function of yaw rate loop can be found by
rejection). using a root locus technique [14] to get the gain value which
damp the rate and make it suitable for maintaining yaw rate
from rudder to be zero. Yaw damper is designed with washout
filter as mentioned above. From the lateral state space
linearized model [4], the linearized transfer function of yaw
rate to rudder was called to be used.

ISBN: 978-1-61804-341-2 106

 Recent advances on Mathematical Models for Engineering Science

() 82.04 3 1385 2 1228 2351 IX. CONCLUSION

() = (5)
4 +19.74 3 +90.49 2 +502 .2 +6.89
In this paper, the design of autopilot is overviewed with the
The lateral model has two pair complex poles (- 1.84 detailed design of lateral motion controller. Most inner loop
5.28i) with a lightly damping ratio (0.329) and natural was designed with feedback gain, and then roll attitude hold
frequency (5.59 rad/sec). The purpose of the damper is to controller was designed with PI-controller far away from
increase this damping ratio. complexity with good performance in the time domain
The controlled system was designed by assigning the gain characteristics. The design procedures began with roll rate
kd_r = 0.065 and the washout filter with transfer function: feedback, the roll attitude controller, Linearization of the

() = (6) nonlinear equation of motion of heading rate is derived to get
+1
Time constant = 0.5 a linear relation between heading angle and roll attitude. The
outer loop controller is a simple P-controller. Yaw damper
VIII. CHECK THE PERFORMANCE OF THE LATERAL MOTION was designed with washout filter. At last a rectangular motion
CONTROLLER FOR SUAV. command is applied in the non-linear model to check the
behavior of the aircraft. The environment disturbances and
The following test; the rectangular motion response in
sensors noise are modeled and the performance of the system
Software In Loop (SIL) which is applied in the non-linear
is checked in the existence of them. Longitudinal motion
model of aircraft, Fig.12 illustrates the response of heading
controller is in the development stage to design the whole
angle due to rectangular motion command, Fig. 13 shows the
relation between the latitude and longitude, and Fig.14 show autonomous SUAV with a robust autopilot.
the roll change (inner loop command) which make the
rectangular motion. X. REFERENCES
Rectangular motion with heading angle psi in SIL
400 [1] B. L. Stevens and F. L. Lewis, Aircraft Control and Simulation, 2nd
350
Ed., Hoboken, NJ: John Wiley & Sons, Inc., 2003.
[2] Gleason, D. Gebre-Egziabher, GNSS Applications and Methods
300
Artech House, Boston, 2009.
Stewart, Calculus: Early Transcendentals, 4th Ed. Cole, Belmont,
heading angle [deg]

250 [3]
200
2001.
[4] Ahmed EA, Hafez A, Ouda AN, Ahmed HEH, Abd-Elkader HM
150
"Modeling of a Small Unmanned Aerial Vehicle" international journal
100 of mechanical, aerospace, industrial and mechatronics engineering Vol:
9, No: 3 413-421 2015.
50
classic controller
designed controller
[5] Reed Siefert Christiansen, Design of An Autopilot for Small
0
0 10 20 30 40 50 60 70 80 Unmanned Aerial Vehicles Brigham young university, august 2004.
time[sec]
[6] Andrei Dorobantu, Peter J. Seiler, Gary J. BALAS, Test Platforms for
Fig. 12 Heading command rectangular motion response. Model-Based flight research, University of Minnesota, 2013.
-1.6244
The trajectory of the plane due to rectangular command in the heading angle [7] G. F. Franklin, J. D. Powell, and M. Workman, Digital Control of
The trajectory of the plane
Dynamic Systems, 3rd Ed. Menlo Park, CA: Addison Wesley, 1998.
-1.6244
[8] Thomas R. Yechout, Introduction to Aircraft Flight Mechanics,
-1.6244
AIAA, 2003.
-1.6244 [9] Peter J. Swatton, Principles of Flight for Pilots, John Wiley & Sons,
-1.6244 June 2011.
[10] Randal W.Beard, Timothy W.Mclain, Small Unmanned Aircraft:
LON

-1.6244

-1.6244
Theory and Practice, Princeton university press, 2012.
-1.6244
[11] Murch, A., Dorobantu, A., and Balas, G., University of Minnesota
UAV Flight Control Research Group, http://www.uav.aem.umn.edu, 8
-1.6245
January, 2014].
-1.6245
[12] W. F. Phillips, Mechanics of Flight, 2nd Ed. New Jersey: Wiley, 2010.
-1.6245
0.7806 0.7806 0.7806 0.7806 0.7806 0.7806 0.7806 0.7806 0.7806
[13] Dongwon Jung, Hierarchical Path Planning and Control a Small Fixed-
LAT Wing UAV: TheoryaAnd Experimental Validation Georgia institute of
Fig.13 The shape of the aircraft trajectory due to rectangular technology, December 2007.
[14] Farid Colnaraghi, Benjamin C.kuo, Automatic Control Systems, 9th Ed.
motion commands in the heading angle. , john wiley&sons, 2010.
Rectangular motion with change in roll in SIL
50
classic controller
designed controller
40
heading angle [deg]

30

20

10

-10
0 10 20 30 40 50 60 70 80
time[sec]

Fig. 14 The change in the bank angle due to rectangular

motion in the aircraft.

ISBN: 978-1-61804-341-2 107