Content uploaded by Andrus Pedai
Author content
All content in this area was uploaded by Andrus Pedai on Jan 28, 2014
Content may be subject to copyright.
Enhancing Situational Awareness through Neural Regulation of
Take-Off and Landing Manoeuvres in Unmanned Helicopter
IGOR ASTROV, ANDRUS PEDAI
Department of Computer Control
Tallinn University of Technology
Ehitajate tee 5, Tallinn 19086
ESTONIA
igor.astrov@dcc.ttu.ee, andrus.pedai @dcc.ttu.ee
Abstract: - This paper focuses on a critical component of the situational awareness, the neural control of
autonomous vertical flight for an unmanned aerial vehicle. Application of the proposed two stage flight
strategy which uses two autonomous adaptive neural dynamical feedback controllers was carried out for a
nontrivial small-scale helicopter model comprising five states and two inputs. This control strategy for chosen
helicopter model has been verified by simulation of take-off and landing manoeuvres using software package
Simulink and demonstrated good performance for fast situational awareness in real-time search-and-rescue
operations.
Key-Words: - Flight control, helicopter, neural networks, simulation, situational awareness.
1 Introduction
Situation awareness has been formally defined as
"the perception of elements in the environment
within a volume of time and space, the
comprehension of their meaning, and the projection
of their status in the near future" [1]. As the term
implies, situation awareness refers to awareness of
the situation. Grammatically, situational awareness
(SA) refers to awareness that only happens
sometimes in certain situations.
SA has been recognized as a critical, yet often
elusive, foundation for successful decision-making
across a broad range of complex and dynamic
systems, including emergency response and military
command and control operations [2].
The term SA have become commonplace for the
doctrine and tactics, and techniques in the U.S.
Army [3]. SA is defined as "the ability to maintain a
constant, clear mental picture of relevant
information and the tactical situation including
friendly and threat situations as well as terrain". SA
allows leaders to avoid surprise, make rapid
decisions, and choose when and where to conduct
engagements, and achieve decisive outcomes.
The Tactical Unmanned Aerial Vehicle (TUAV)
is one of the key tools to gather the information to
build SA for all leaders. The TUAV is the ground
maneuver commander's primary day and night
system.
The TUAV provides the commander with a
number of capabilities including:
Enhanced SA.
Target acquisition.
Battle damage assessment.
Enhanced battle management capabilities
(friendly situation and battlefield visualization).
The combination of these benefits contributes to
the commander's dominant SA allowing him to
shape the battlefield to ensure mission success and
to maneuver to points of positional advantage with
speed and precision to conduct decisive operations.
Some conditions for conducting aerial
reconnaissance with TUAVs are as follows.
Time is limited or information is required
quickly.
Detailed reconnaissance is not required.
Extended duration surveillance is not required.
Objective is at extended range.
Verification of a target is needed.
Threat conditions are known and risk to ground
assets is high.
Terrain restricts approach by ground units.
Terrain and weather conditions are favorable.
A small-scale unmanned helicopter offers many
advantages, including low weight and cost, the
ability to fly within a narrow space and the unique
hovering and vertical take-off and landing (VTOL)
flying characteristics.
Autonomous vertical flight is a challenging but
important task for TUAVs to achieve high level of
autonomy under adverse conditions. The
fundamental requirement for vertical flight is the
knowledge of the height above the ground, and a
properly designed controller to govern the process.
Proc. of the 7th WSEAS Int. Conf. on COMPUTATIONAL INTELLIGENCE, MAN-MACHINE SYSTEMS and CYBERNETICS (CIMMACS '08)
ISSN: 1790-5117
29
ISBN: 978-960-474-049-9
This paper presents our research results in the
study of vertical flight (both take-off and landing)
neural control of autonomous unmanned small-scale
helicopters which make such SA task scenario as
"go-search-find-return" possible.
With the SA strategy, we proposed a two stage
flight control procedure using two ADAptive LInear
NEuron neural networks (ADALINE NNs) to
address the dynamics variation and performance
requirement difference in initial and final stages of
flight trajectory for an unmanned small-scale
helicopter.
The contribution of the paper is twofold: to
develop new schemes appropriate for SA
enhancement using TUAVs by neural network
control of vertical flight of autonomous unmanned
small-scale helicopters in real-time search-and-
rescue operations, and to present the results of take-
off and landing manoeuvres for chosen model of the
helicopter for fast SA in simulation form using the
MATLAB/Simulink environment.
2 ADALINE
ADALINE was developed in [4]. It consists of a
weight, a bias and a summation function.
Operation:
bwxy ii += (1)
Its adaptation is defined through a cost function
(error metric) of the residual
()
bwxde ii +−= (2)
where i
d is the desired signal.
With the mean squared error metric
∑
=N
ii
e
N
E2
2
1 (3)
the bias and adapted weight become:
()
∑
∑∑ ∑∑
−
−
=
ii
ii ii
iiiii
xxN
dxxdx
b2
2
(4)
(
)
(
)
()
∑
∑
−
−−
=
ii
iii
xx
ddxx
w2 (5)
3 Helicopter Model
Consider the nonlinear helicopter model [5] in terms
of a state variable representation as follows:
(
)
(
)()
2211 uxguxgxfx +
+
=
& (6)
Cxy
=
(7)
where
()
()
()()
,
sin
sin
5144
2
31341211
5
2
31049387
2
3465443
2
22210
2
+++
+++
+−++++
=
xaxxaxaa
x
xaxaxaa
xxaaxaaxaxaa
x
xf
() ()
,
1
0
0
0
0
,
0
0
1
0
0
21
=
=xgxg
,
01000
00001
=C
[
]
,
21
T
uuu=
[
]
,
T
hhx
θθω
&
&
=
[
]
,
21
T
yyy=
=
hheight above ground (m),
=
ω
rotational speed of the rotor blades (radn/s),
=
θ
collective pitch angle of rotor blades (radn),
=
1
uinput to the throttle,
=
2
uinput to collective servomechanisms.
The parameters 0
a through 14
a are given by:
,/67.17 2
0sma −= ,1.0 1
21 −
−== saa ,1031.5 4
3−
×=a
,105364.1 2
4−
×=a,1082.2 7
5−
×=a,10632.1 5
6−
×=a
,92.13 2
7−
−= sa ,7.0 1
8−
−= sa ,0028.0
109
−
== aa
,88.434 2
11 −
=sa ,800 2
12 −
−= sa 1.0
13 −=a and
1
14 65 −
−= sa .
4 Simulation
Since the height is most critical for take-off and
landing manoeuvres, the control mechanization of
the vertical trajectory profile will be demonstrated.
To illustrate the performance of the neural
control design procedure for the helicopter model
given by (6)-(7), we present three simulation
examples: the first one addresses to control the take-
off and landing trajectory by one neural adaptive
controller, while the second example presents the
Proc. of the 7th WSEAS Int. Conf. on COMPUTATIONAL INTELLIGENCE, MAN-MACHINE SYSTEMS and CYBERNETICS (CIMMACS '08)
ISSN: 1790-5117
30
ISBN: 978-960-474-049-9
case of two neural adaptive controllers using to
control the take-off and landing trajectory, and the
third example shows the case of control for take-off
and hovering trajectory by hybrid system with two
neural adaptive controllers.
The goal of the following simulations is twofold.
First, we verify that these neural adaptive controllers
are able to control the take-off and landing
trajectory. Second, we observed the effect of
enhancing SA because the variety of such trajectory
parameters as maximal height of flight and height of
hovering above a point of a landing easily can be
changed the possible take-off and landing trajectory
of helicopter.
4.1 Example 1
Consider the control of helicopter model (6)-(7) by
one neural adaptive controller.
Initial conditions are chosen to be:
,0)0(
1mx =,/0)0(
2smx =,/200)0(
3sradnx =
,15.0)0(
4radnx =./0)0(
5sradnx =
Simulation results of the offered block scheme
with one neural adaptive controller (see Fig. 1) are
given in Fig. 6.
In1
In2
Out1
Out2
System
Step u2
Step u1
Scope2
Scope1
p{1}
p{2}
y{1}
y{2}
ADALINE NN
Fig. 1. Block diagram of system with one neural
adaptive controller.
A structure of the helicopter model (6)-(7) is
illustrated in Fig. 2.
Detailed block diagrams of the ADALINE NN
from Fig. 1 are given in Figs. 3-5.
2
Out2
1
Out1
sin
Trigonometric
Functi on
Product2
Product1
Product
sqrt
Math
Functi on2
u2
Math
Function1
u2
Math
Functi on
1/s
Integrator 5
1/s
Integrator 4
1/s
Integrator 3
1/s
Integrator 2
1/s
Integrator 1
C* u
Gain9
a6
Gain8
a4
Gain7
a2 Gain6
a1 Gain5
a9 Gain4
a8
Gain3
a14
Gain2
a13
Gain1
a12
Gain
a7
Constant7
a5
Constant5
a3
Constant3
a11
Constant11
a10
Constant10
a0
Constant0
2
In2
1
In1
Fig. 2. The internal structure of the System from
Fig. 1.
2
y{2}
1
y{1}
p{1}
p{2} a{2}
Layer 2
p{1}
p{2} a{1}
Layer 12
p{2}
1
p{1}
Fig. 3. Block diagram of ADALINE NN from
Fig. 1.
1
a{1}
purelin
netsum
w
pz
dotprod2
w
pz
dotprod1
bias
b{1}
Mux
Mux2
Mux
Mux1
weights
IW{1,2}(1,:)'
wei ghts
IW{1,1}(1,:)'
TDL
Delays 2
TDL
Delays 1
2
p{2}
1
p{1}
Fig. 4. Block diagrams of Layer 1 from Fig. 3.
Proc. of the 7th WSEAS Int. Conf. on COMPUTATIONAL INTELLIGENCE, MAN-MACHINE SYSTEMS and CYBERNETICS (CIMMACS '08)
ISSN: 1790-5117
31
ISBN: 978-960-474-049-9
1
a{2}
purelin
netsum
w
pz
dotprod2
w
pz
dotprod1
bias
b{2}
Mux
Mux2
Mux
Mux1
wei gh ts
IW{2,2}(1,:)'
wei gh ts
IW{2,1}(1,:)'
TDL
Delays 2
TDL
Delays 1
2
p{2}
1
p{1}
Fig. 5. Block diagrams of Layer 2 from Fig, 3.
00.2 0.4 0.6 0.8 11.2 1.4 1.6 1.8
0
0.5
1
1.5
2
2.5
3
Time (s )
Height (m)
Height trajectory
Fig. 6. Height trajectory of flight control using one
neural adaptive controller during take-off and
landing manoeuvre.
Some disadvantages of this example are as
follows.
Vertical velocity of a helicopter increases as
approaching a place of a landing that can lead to
accident and loss of this helicopter.
Impossibility to consider a terrain restriction in a
place of a landing can lead to accident and loss of
this helicopter.
4.2 Example 2
Consider the control of helicopter model (6)-(7) by
two neural adaptive controllers.
Simulation results of the offered block scheme
with two neural adaptive controllers (see Fig. 7) are
shown in Fig. 8. We simulated the block diagrams of
Systems and ADALINE NNs with the same
structural units given in example 1 except that the
full take-off and landing trajectory was separated
into initial and final phases with boundary point in
the maximal height position.
In1
In2
Out1
Out2
System 2
In1
In2
Out1
Out2
System 1
Switch
Step u4
Step u3
Step u2
Step u1
Scope4
Scope3
Scope2
Scope1
Scope
p{1}
p{2}
y{1}
y{2}
ADALIN E NN 2
p{1}
p{2}
y{1}
y{2}
ADALINE NN 1
Fig. 7. Block diagram of two systems with two
neural adaptive controllers.
00.2 0.4 0.6 0.8 11.2 1.4
0
1
2
3
4
5
6
Time (s)
Height (m )
Height trajectory
Fig. 8. Height trajectory of flight control using two
neural adaptive controllers during take-off and
landing manoeuvre.
Proc. of the 7th WSEAS Int. Conf. on COMPUTATIONAL INTELLIGENCE, MAN-MACHINE SYSTEMS and CYBERNETICS (CIMMACS '08)
ISSN: 1790-5117
32
ISBN: 978-960-474-049-9
Some advantages of this example are as follows.
Possibility to consider a terrain restriction in a
place of a landing.
Possibility of hovering above a place of a
landing.
Minimization of flight time.
Using of two simple adaptive NNs to control the
take-off and landing trajectory.
4.3 Example 3
Consider the control of helicopter model (6)-(7) for
the case of take-off and hovering trajectory by
hybrid system of two neural adaptive controllers.
Simulation results of the offered block scheme
with two neural adaptive controllers (see Fig. 9) are
shown in Fig. 10. We simulated the block diagrams
of Systems and ADALINE NNs with the same
structural units given in example 1 except that the
full take-off and hovering trajectory was separated
into initial and final phases with boundary point in
the maximal height position.
Transport
Delay
In1
In2
Out1
Out2
System 2
In1
In2
Out1
Out2
System 1
Switch3
Switch2
Switch1
Step u3
Step u2
Step u1
Scope4
Scope3
Scope2
Scope1
Scope
Saturation2
Saturation1
12:34
Digi tal Clock
p{1}
p{2}
y{1}
y{2}
ADALIN E NN 2
p{1}
p{2}
y{1}
y{2}
ADALIN E NN 1
Fig. 9. Block diagram of hybrid system with two
neural adaptive controllers.
Some advantages of this example are as follows.
Opportunity of smooth switching of regulation
from one controller to another (see Fig. 11).
Possibility to consider a terrain restriction in a
place of a hovering (see Fig. 12).
Smoother trajectory of flight in comparison with
examples 1-2.
Possibility of lag in the maximal height position.
Use of one general operating input (input to the
throttle) does more simple regulation for this hybrid
scheme.
Using of two simple adaptive NNs to control the
take-off and hovering trajectory.
00.5 11.5 22.5 3
0
1
2
3
4
5
6
7
8
9
10
Time (s)
Height (m)
Height trajectory
Fig. 10. Height trajectory of flight control using two
neural adaptive controllers during take-off with
hovering manoeuvre.
1.4 1.45 1.5 1.55 1.6 1.65 1.7 1.75 1.8
8.6
8.8
9
9.2
9.4
9.6
9.8
10
Tim e (s )
Height (m)
Height trajec tory
Fig. 11. Height trajectory near the maximal position.
Proc. of the 7th WSEAS Int. Conf. on COMPUTATIONAL INTELLIGENCE, MAN-MACHINE SYSTEMS and CYBERNETICS (CIMMACS '08)
ISSN: 1790-5117
33
ISBN: 978-960-474-049-9
2.6 2.62 2.64 2.66 2.68 2.7 2.72 2.74 2.76
1.9
2
2.1
2.2
2.3
2.4
Tim e (s)
Height (m)
Height trajectory
Fig. 12. Height trajectory near the hovering position.
These results support the theoretical predictions
well and demonstrate that this research technique
would work in real-time flight conditions.
5 Conclusions
The need for take-off capability and highly reliable
and stable landing for helicopters and VTOL class
autonomous vehicles has increased morbidly for
critical situations in real-time search-and-rescue
operations for fast SA.
For fast and stable take-off and smooth landing
manoeuvres, we proposed a two stage flight
strategy, which separates the flight process into
initial and final phases. Two controllers on the base
of ADALINE NNs as feedback gain controllers are
designed for the initial phase and final phase
respectively. The effectiveness of the proposed two
stage flight strategy has been verified in field of
flight simulation tests for chosen model of the
helicopter using software package Simulink.
From the simulation studies of flight tests, the
following can be observed:
The block diagram of flight neural control is very
useful for graphic representation of the flight
trajectory.
The received controllers for various flight phases
are autonomous and completely shared in time.
The trajectory tracking display forms give a
researcher an immediate view of a helicopter motion
with a range of such trajectory parameters as
maximal height of flight and height of hovering
above a place of a landing. This allows us to
investigate the sensitivity of the control system,
providing a medium for such development and
evaluation and enhancing the researcher’s
understanding of take-off and landing manoeuvres.
The neural control using two stage flight
strategies works more quickly and qualitatively than
the neural control using one stage flight strategy.
From the applications viewpoint, we believe that
this two stage flight strategy using flexible and
effective neural control furnish a powerful approach
for enhancing SA in applications to VTOL class
autonomous vehicles.
Recently, a new flight strategy [6] has been
developed for descending control using two general
operating inputs of helicopter model.
In the future, we will extend the proposed
research technique to other types of manoeuvres, by
vertical flight simulation for a model helicopter
using the MATLAB/Simulink software.
References:
[1] M. R. Endsley, Toward a Theory of Situation
Awareness in Dynamic Systems, Human
Factors, Vol. 37, No. 1, 1995, pp. 32-64.
[2] J. Gorman, N. Cooke and J. Winner, Measuring
Team Situation Awareness in Decentralized
Command and Control Environments,
Ergonomics, Vol. 49, Nos. 12-13, 2006, pp.
1312–1325.
[3] Interim Brigade Combat Team Newsletter.
[Online]. Available: http://www.globalsecurity.
org/ military/ library/ report/ call/call_01-18_toc.
htm
[4] B. Widrow and S.D. Sterns, Adaptive Signal
Processing, New York: Prentice-Hall, 1985.
[5] J. Kaloust, C. Ham and Z. Qu, Nonlinear
Autopilot Control Design for a 2-DOF
Helicopter Model, IEEE Proc.-Control Theory
Appl., Vol. 144, No. 6, 1997, pp. 612-616
[6] I. Astrov and A. Pedai, Enhancing Situational
Awareness by means of Hybrid Adaptive Neural
Control of Vertical Flight in Unmanned
Helicopter, USB Proc. International Conf.
Control, Automation and Systems, Seoul, Korea,
2008, pp. 329-332.
Proc. of the 7th WSEAS Int. Conf. on COMPUTATIONAL INTELLIGENCE, MAN-MACHINE SYSTEMS and CYBERNETICS (CIMMACS '08)
ISSN: 1790-5117
34
ISBN: 978-960-474-049-9
