You are on page 1of 10

Corresponding author: Xiao-lei Yu

E-mail: xiaoleiyu@nuaa.edu.cn

Journal of Bionic Engineering 6 (2009) 270279


Autonomous Navigation for Unmanned Aerial Vehicles
Based on Chaotic Bionics Theory

Xiao-lei Yu
1,2
, Yong-rong Sun
1
, Jian-ye Liu
1
, Bing-wen Chen
3

1. Navigation Research Centre, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, P. R. China
2. Institute for Technology, Research & Innovation, Deakin University, Vic 3217, Australia
3. Institute for Unmanned Aerial Vehicle Research, Beihang University, Beijing 100083, P. R. China

Abstract
In this paper a new reactive mechanism based on perception-action bionics for multi-sensory integration applied to Un-
manned Aerial Vehicles (UAVs) navigation is proposed. The strategy is inspired by the olfactory bulb neural activity observed in
rabbits subject to external stimuli. The new UAV navigation technique exploits the use of a multiscroll chaotic system which is
able to be controlled in real-time towards less complex orbits, like periodic orbits or equilibrium points, considered as perceptive
orbits. These are subject to real-time modifications on the basis of environment changes acquired through a Synthetic Aperture
Radar (SAR) sensory system. The mathematical details of the approach are given including simulation results in a virtual en-
vironment. The results demonstrate the capability of autonomous navigation for UAV based on chaotic bionics theory in com-
plex spatial environments.
Keywords: chaotic system, perception-action bionics, UAV, multi-sensory integration, autonomous navigation
Copyright 2009, Jilin University. Published by Elsevier Limited and Science Press. All rights reserved.
doi: 10.1016/S1672-6529(08)60123-7

1 Introduction
Unmanned Aerial Vehicles (UAVs) have become
the research focus in recent years because of their broad
application future both in military affairs and in civil
aspects
[1,2]
. In military affairs, UAVs are able to fight and
reconnoitre in very dangerous zones in order to reduce
the injuries and death of people. On the other hand,
UAVs can be used to achieve heavy and repeated tasks
such as resource surveys, observation of the situation of
disaster, communication transfer, environment moni-
toring and so on. The study on autonomous navigation of
UAVs is a frontier problem of flight control. The goal is
to realize the autonomous flight control, decision mak-
ing and management for UAVs. Because of the huge
complexity and high intelligence of autonomous navi-
gation of UAVs, the research is still in the early stage.
Under traditional control strategies, the navigation
of UAVs can be implemented by close-distance com-
munication chains constructed by other formation of
aircrafts. In longer distance, the navigation of UAVs can
be controlled by command platform either on the ground
or in the sky. In addition, the satellite communication
can be used to assist the navigation of UAV
[3]
. However,
all the methods reviewed above adopted the way of
controlling the UAVs by external data communication
chains. In an atrocious environment, the result will be
unpredictable if the communication chain is neither
credible nor smooth. Therefore, the ability of inde-
pendence, self-adaptability and self-determination is the
key of the new generation of UAVs.
Verschure and his co-workers developed a percep-
tual scheme, called Distributed Adaptive Control (DAC),
as a neural model for classical and operant condition-
ing
[4,5]
. In DAC three tightly connected control layers are
introduced: the reactive layer, the adaptive layer and the
contextual layer. The reactive control layer implements a
set of reflexes based on low level sensorial uncondi-
tioned inputs. The adaptive control layer allows the
system to associate more complex stimuli with the basic
ones. The contextual layer constructs high-level repre-
sentations by means of memory structures.
All forms of adaptive hehaviors require the proc-
essing of multi-sensory information and their transfor-

Yu et al.: Autonomous Navigation for Unmanned Aerial Vehicles Based on Chaotic Bionics Theory
271
mation into a series of goal-directed actions. In the most
primitive animal species the entire process is regulated
by external (environmental) and internal feedback
through the animal body. Freeman and his co-workers,
in their long experimental studies on the bionics of
sensory processing in animals, conceived a dynamic
theory of perception
[610]
. The hypothesis is that cerebral
activity can be represented by chaotic dynamics. They
attained this result by different experiments on rabbits
which inhaled several smells in a pre-programmed way.
Through the electroencephalogram (EEG), Freeman
evaluated the action potentials in the olfactory bulb and
noticed that the potential waves showed a complex he-
havior. So he concluded that an internal representation
(cerebral pattern) of a stimulus is the result of complex
dynamics in the sensory cortex in cooperation with the
limbic system that implements the supporting processes
of intention and attention. In more detail, the dynamics
of the olfactory bulb is characterized by a high-dimen-
sional chaotic attractor with multiple scrolls. The scrolls
can be considered as potential memory traces formed by
learning through the animals life history. In the absence
of sensory stimuli, the system is in a high dimensional
itinerant search mode, visiting various scrolls. In re-
sponse to a given stimulus, the dynamics of the system is
constrained to oscillate in one of the scrolls, which is
identified with the stimulus. Once the input is removed,
the system switches back to the high-dimensional, itin-
erant basal mode. According to their works, neural sys-
tem activity persists in a chaotic state until sensors per-
turb this hehavior. The result of this process is that a new
attractor emerges representing the meaning of the in-
coming stimuli. The role of chaos is fundamental to
provide the flexibility and the robustness needed by the
system during the migration through different perceptual
states. A discrete implementation of Freemans model
was developed and applied to navigation control of
autonomous agents
[11]
. The controller parameters have
been learned through an evolutionary approach and also
by using unsupervised learning strategies. Furthermore,
some work on Knowledge-oriented Information Visu-
alization (KIV) sets and robotics applications referring
to Freemans model have been presented in recent
years
[1215]
.
In this paper, taking inspiration from Freemans
experiments showing the presence of chaotic dynamics
in neural system activities, and paying attention to the
reactive layer in DAC, we adopt a new bionic control
architecture to implement the perception-action loop.
We apply the proposed architecture to the navigation
control problem for UAV. We do not try to reproduce a
bio-relevant model of neural assemblies; our main ob-
jective is to propose reactive control architectures for
autonomous UAVs, taking care of the functional prop-
erties discovered by Freeman in the olfactory bulb. The
idea is to use a simple but chaotic dynamic system with
suitable characteristics that can functionally simulate the
creation of perceptual patterns. The bionic patterns can
be used to guide the UAVs actions and the control sys-
tem can be easily extended to include a large number of
sensors.
2 Bionic evolution theory of the multiscroll
chaotic system
Recently, multiscroll chaotic technique has motived
many scientists from different disciplines. From the
application point of view, the double scroll attractor has
been successfully employed towards true random bit
generation
[16]
. Furthermore, the design, generation and
analysis of multiscroll chaotic attractors become the
leading edge gradually
[1720]
. Since this UAV system
should be able to deal with a great number of sensorial
stimuli and represent them, a chaotic system, able to
generate multiscrolls, has been adopted
[17]
. This can
generate a chaotic attractor consisting of multiple scrolls
distributed in the phase space. In this paper a
three-dimensional (3D) multiscroll system is chosen,
which is able to generate 3D n m l grid scroll chaotic
attractors by using saturated function series. It is de-
scribed by the following linear differential equations:
2
2 2 2 2
3
3 3 3 3
1 1 1 1 1
2 2 2 2 2 3 3 3 3 3
( , , , , )
( , , , , )
( , , , , )
( , , , , ) ( , , , , )
x y f y k h p q
b
y z f z k h p q
c
z ax by cz f x k h p q
f y k h p q f z k h p q

= + +

, (1)
where the following so-called saturated function series
f(x,k,h,p,q) is used:
( )
( )
( )
( )
2 1 , 1
( ) 2 , 1,
, , , ,
2 1 , 1
2 1 , 1
q k x qh
k x ih ik x ih p i q
f x k h p q
i k p i q
p k x ph
+ > +

+ s s s

+ s s

+ <

(2)

Journal of Bionic Engineering (2009) Vol.6 No.3
272
The system represented by Eq. (1) can generate a
grid of (p
1
+q
1
+2) (p
2
+q
2
+2) (p
3
+q
3
+2) scroll at-
tractors. Parameters (p
1
,q
1
), (p
2
,q
2
), (p
3
,q
3
) control the
number of scroll attractors in the direction of variables x,
y, z, respectively. In Eq. (1) the coefficients a, b, c,
1
,
2

and
3
are all positive constants. In particular, coeffi-
cients a, b, c control the shape and area of the limit cycle
scroll. Coefficients
1
,
2
,
3
control the convergence
during the evolution of the multiscroll system when
controlled by reference dynamics associated with sen-
sors. The parameters used in Eq. (3) have been chosen
according to the guidelines introduced by Lu et al. to
generate a 3D 6 6 6 grid of scroll attractors
[17]
. Under
these conditions, the saturated function series (Eq. (2)) is
given in Fig. 1, and an example of the chaotic dynamics
of system (Eq. (1)) is given in Fig. 2.
1 2 3 1
2 3 1 2 3
1 2 3 1 2 3
0.7, 0.8, 0.7, 0.8, 100,
40, 200, 80,
2.
a b c k
k k h h h
p p p q q q
= = = = = = =
= = = = =
= = = = = =

(3)

600 400 200 0 200 400 600
600
400
200
0
200
400
600
x
f

(
x
)

f
(
x
)

(a) k = 100, h = 200, p = q = 2 (b) k = 40, h = 80, p = q = 2
Fig. 1 Saturated function series.


Fig. 2 Projection of the 6 6 6 grid of scroll attractors in the x-y-z space.

Yu et al.: Autonomous Navigation for Unmanned Aerial Vehicles Based on Chaotic Bionics Theory
273

In our approach the perceptual system is repre-
sented by the multiscroll attractors in Eq. (1), whereas
sensorial stimuli can interact with the system through
constant or periodic inputs that can modify the internal
chaotic behavior. Since one of the main characteristics of
perceptive systems is that sensorial stimuli strongly
influence the spatial-temporal dynamics of the internal
state, a suitable scheme to control the chaotic behavior
of the multiscroll system on the basis of sensorial stimuli
should be adopted. Briefly, chaos control refers to a
process wherein a tiny perturbation is applied to a cha-
otic system in order to realize a desirable behavior (e.g.
chaotic, periodic and others). Several techniques have
been developed for the control of chaos. Traditionally,
chaotic control often means converting or constraining
chaos to other dynamics, however, what we are adopting
here is a broader and more general approach, and more
powerful. In view of our application, a continuous-time
technique like the Pyragass method is a suitable
choice
[21]
. In the method, the following model is taken
into account:
( ) ( ) ( )
, F t k y t y t = (

(4)
where F(t) is the feedback perturbation which forces the
chaotic system to follow the desired dynamics, that is,
the stabilization of unstable periodic orbits of a chaotic
system is achieved by combining feedback with the use
of a specially designed external stimulus; ( )
y t repre-
sents the external input (i.e. the desired dynamics); and k
represents a vector of experimental adjustable weights
(adaptive control). The method can be employed to sta-
bilize the unstable orbits in the chaotic attractor, reduc-
ing the high order dynamics of a chaotic system.
The desired dynamics is provided by a constant or
periodic signal associated with more than one sensorial
stimulus. Hence, the equations of the controlled multis-
croll system can be written as follows:
( ) ( )
( ) ( )
( )
( ) ( )
( )
2
2 2 2 2
3
3 3 3 3
1 1 1 1 1
2 2 2 2 2 3 3 3 3 3
, , , ,
, , , ,
, , , ,
, , , , , , , ,
xi si
i
yi si
i
zi si
i
x y f y k h p q k x x
b
y z f z k h p q k y y
c
z ax by cz f x k h p q
f y k h p q f z k h p q
k z z

= +

= +

= + +

+ +

, (5)
where i is the number of external stimulus acting on the
system, x
si
, y
si
and z
si
are the state variables of the
stimulus signals that will be described in details below in
Eq. (6), and k
xi
, k
yi
and k
zi
represent the control gains in
the direction of x, y and z respectively. Each stimulus
signal (x
si
, y
si
, z
si
) can be a constant input or a periodic
trajectory representing a native cycle. In a more simple
way, these stimulus signals can be built using sinusoidal
oscillators:
( ) ( )
( ) ( )
( ) ( )
center
center
center
sin
sin
sin
si xsi xsi xsi
si ysi ysi ysi
si zsi zsi zsi
x t A t x
y t A t y
z t A t z
e
e
e
= + +

= + +

= + +

,(6)
where ( )
center center center
, , x y z is the centre of the stimulus
cycle, is the frequency, is the phase, and A defines
the amplitude of the stimulus signal. The geometry of
stimulus cycle can be adjusted by changing these pa-
rameters above.
The bionic evolution of the multiscroll chaotic
system is closely related to the stimulus. When no
stimuli are added to the system, the hehavior of chaotic
dynamics is shown in Fig. 3a. When stimuli are per-
ceived, the system immediately converges to a limit
cycle that constitutes a representation of the concurrent
activation of the sensorial stimuli, as seen in Fig. 3b.
When the stimuli are perceived consistently, the system
will be constrained to oscillate in one of the scrolls, as
shown in Fig. 3c. When stimuli are no longer active, the
multiscroll returns to its default chaotic dynamics, and
this process is shown in Fig. 3d.

800
600
400
200
0
200
400
600
800
400
200
0
200
400
400
300
200
100
0
100
200
300
400
X
Y
Z
(a)
Fig. 3 An example of the bionic evolution of the multiscroll
system when controlled by stimuli dynamics.

Journal of Bionic Engineering (2009) Vol.6 No.3
274

800
600
400
200
0
200
400
600
800
400
200
0
200
400
400
300
200
100
100
200
300
400
X
Y
0

(b)

Z

(c)

Z

(d)
Fig. 3 Continued.

3 Autonomous navigation of UAV
When a UAV is placed in an unknown environment,
it is subject to a huge amount of external stimuli. To
explore the area avoiding obstacles, the UAV, sensing
the environment, can create an internal representation of
the stimuli in relation to its body. The loop is closed by
an action that is chosen to accomplish a given target
behavior (e.g. exploration with obstacle avoidance). In
the previous section we have defined the bionic evolu-
tion theory based on a dynamic system involved in re-
alizing a reactive navigation layer. It is based on a cha-
otic control technique, used to enslave the dynamics of a
multiscroll attractor to follow one or more trajectories.
In this paper, each sensor equipped on the UAV
provides a stimulus cycle. This is addressed by associ-
ating, for instance, the perception of an obstacle with a
stimulus and associating it with a representation (pat-
tern). Therefore the controlled multiscroll system is the
perceptual system and the emerged orbit stands for the
internal representation of the external environment.
Finally, according to the characteristic of the emerged
cycle (amplitude, frequency, centre position), an action,
in terms of rotation angle, is associated. In this bionic
strategy, action is linked to perception (the emerged
cycle) using a deterministic algorithm. However, this
association can be obtained through a bio-inspired
adaptive structure which is suitable to control the UAV
navigation in an unknown environment because it is
adaptive and unsupervised. The adaptive structure is
called the perception-action dynamics loop shown in Fig.
4. In the structure, the multiscroll chaotic reaction sys-
tem can be described by Eq. (1), while the controlled
multiscroll chaotic system with state feedback is pre-
sented by Eq. (5), and the actuator refers to the steering
engine of UAV, associated with rotation angle.


Fig. 4 Structure of perception-action dynamics loop.

Synthetic Aperture Radar (SAR), with high resolu-
tion, was widely used in some military fields. In our
work, SAR sensor is chosen as a good candidate placed
on the UAV. Among the possible stimulus cycles which
can be used, in the application here reported, their dis-
tribution in the 3D phase space reflects the topological
distribution of UAV sensors. Fig. 5 shows a scheme of

Yu et al.: Autonomous Navigation for Unmanned Aerial Vehicles Based on Chaotic Bionics Theory
275
the UAV equipped with a SAR sensor. The correspond-
ing stimulus cycles in the phase space are related to the
sensor positions. Since SAR sensor is characterized by
an omnidirectional field of view, it is associated with
more than one stimulus cycle. Surely, the stimulus cy-
cles are distributed in 3D space symmetrically. The
offsets assigned to each input cycle are defined to match
the position of the scroll centers. Moreover we choose to
link the value of the control gains with the intensity of
perceived sensorial stimuli. The technique, based on
placing stimulus cycles in the phase space in accordance
with the distribution of sensors on the UAV, is important
to strictly connect the internal representation of the en-
vironment with the UAV geometry.


z
x
Stimuli cycle
y

Fig. 5 Scheme of the UAV equipped with a SAR sensor.

The action (in terms of absolute value of heading)
performed by the UAV depends on the multiscroll be-
havior. In particular, in the absence of constraints the
system evolves chaotically. Also, it can be chaotic even
with constraints, just different types. Moreover, the
possible exploring strategy taken into consideration is
that the UAV continues to explore the environment fly-
ing without modifying its orientation. When external
stimuli are perceived, the controlled system converges to
a cycle (i.e. a periodic pattern) that depends on the con-
tribution of active stimulus through the control gains
from different directions k
xi
, k
yi
and k
zi
. The action to be
executed is chosen according to the characteristics of the
cycle, in particular its centre position in the phase space.
A vector pointing to the center of the limit cycle of the
controlled multiscroll attractor is defined. Predefined
actions are chosen on the basis of the orientation of this
vector. When the stimuli stop, the multiscroll returns to
evolve in a chaotic way. Fig. 6 describes the block dia-
gram of the control scheme when two distinct stimulus
signals (i.e. sensorial stimuli) are perceived by the mul-
tiscroll system.
4 Simulation results
In the following simulations, the parameters are
given in Eq. (3), and the controlled multiscroll chaotic
reaction system is represented by Eq. (5). Essentially, the
selection of control gains k
xi
, k
yi
and k
zi
is very significant,
in particular, taking into account a single stimulus signal,
for low values of k
xi
, k
yi
and k
zi
(as shown in Fig. 7a only
in x-y plane), the control of multiscroll attractor has a
residual error, however for the cases as weak signals
respecting the remote targets, the navigation control is

Multiscroll
chaotic reaction
system
Stimulus 2
k
x1
X control
Y control
Z control
x
y
z
+
+
+
+
+
+
k
y1
k
z1
k
x2
k
y2
k
z2
+

x
s2
y
s2
z
s2
Stimulus 1
x
s1
y
s1
z
s1
x
s1
x
y
s1
y
z
s1
z
x
s2
x
y
s2
y
z
s2
z
+


Fig. 6 Block diagram of the control scheme when two distinct stimulus signals are perceived.

Journal of Bionic Engineering (2009) Vol.6 No.3
276

still acceptable and in accordance with practice. For
higher values of k
xi
, k
yi
and k
zi
(as shown in Fig. 7b only
in x-y plane), the steady error becomes smaller.

Y

(a) Low control gains k
x
= k
y
= k
z
= 0.3

Y

(b) High control gains k
x
= k
y
= k
z
= 1.3
Fig. 7 Behavior of the constrained multiscroll system with dif-
ferent control gains.

In the following section the proposed reactive
control scheme will be applied to UAV navigation con-
trol. The SAR sensor is supposed to be located in the
origin of the state space (i.e. (0,0,0)) in the following
simulations. If the centre of stimulus signal cy-
cle ( )
center center center
, , x y z is (250, 150,150), k
x
= k
y
= k
z
=
1.3, A = 30, = 1, = 0, the equation of the controlled
multiscroll system are adopted as Eq. (5), and the pa-
rameters in the equation are listed in Eq. (3). So when
the single stimulus is perceived, the system immediately
converges to the cycle whose centre is (250, 150, 150)
(as shown in Fig. 8). Again, for example, let us consider
the case in which there are two concurrently active in-
puts, and so there are two stimulus signals in the phase
space. If the two stimulus dynamics have the same con-
trol gains, the resulting cycle will be placed exactly
halfway between them. This result is shown in Fig. 9. In
this case, the centre of one stimulus signal cycle is
(250,150,150), and the centre of another one is
(250,150,150). k
1x
= k
1y
= k
1z
= k
2x
= k
2y
= k
2z
= 1.3, A
1
=
A
2
= 30,
1
=
2
= 1,
1
=
2
= 0. So the centre of
emerged cycle is (250,150,0). If the control gains of the
two stimulus systems are not equal, the resulting con-
trolled limit cycle (emerged cycle) will be placed, in the
phase space, near the stimulus cycle associated with the
higher control gain. This result is shown in Fig. 10. In
this case, the centre of one stimulus signal cycle is
(250,150,150), and the centre of another one is (250,150,
150). k
1x
= k
1y
= k
1z
= 1.2, k
2x
= k
2y
= k
2z
= 2.4, A
1
= A
2
=
30,
1
=
2
= 1,
1
=
2
= 0. So the centre of emerged
cycle is (250,150, 50).

300
200
100
0
100
200
300
300
200
100
0
100
200
300
300
200
100
0
100
200
300
X
Y
Z
Fig. 8 Behavior of the reactive system of UAV controlled by a
single stimulus cycle.

300
200
100
0
100
200
300
300
200
100
0
100
200
300
300
200
100
0
100
200
300
X
Y
Z
Fig. 9 Behavior of the reactive system of UAV controlled by two
equal stimulus cycles.

Yu et al.: Autonomous Navigation for Unmanned Aerial Vehicles Based on Chaotic Bionics Theory
277

300
200
100
0
100
200
300
300
200
100
0
100
200
300
300
200
100
0
100
200
300
X
Y
Z
Fig. 10 Behavior of the reactive system of UAV controlled by two
unequal stimulus cycles.

To test the performance and the potential impact of
the proposed perception-action dynamics loop we de-
veloped a 3D virtual simulation environment. In this
environment, obstacles and no-fly zones are modelled as
cylinders. A small UAV involved in the task of obstacle
avoidance as well as surveillance of several buildings is
planned to go along the trajectories considering the
perceived stimulus. Three different trajectories were
determined by different initial rotation angles associated
with the steering engine of the UAV. Simulation results
in Fig. 11 show the trajectories of the UAV for surveil-
lance of several buildings (represented by four large
cylinders) followed by chaotic control strategy. From the
results we can see if the UAV wants to fly over the four
buildings, it must avoid obstacles from z-direction,
meaning that the system must perceive stimulus from
z-direction and take effective action. On the other hand,
however, the system must perceive stimuli from
x-direction and y-direction and take effective action to
avoid obstacles if the UAV chooses the route that passes
around the buildings. So from the simulation we can
conclude that the perception-action dynamics system has
enough flexibility and robustness in 3D space. Since the
explored area and cumulative number of detected targets
are two key parameters in evaluating the performance of
a robot moving hehavior
[11]
, here we can try some
changes to evaluate the performance of UAV navigation
hehavior in 3D space. We choose the cumulative number
of obstacles found in an environment calculated in time
windows of 10000 actions (i.e. epochs), as well as the
volume explored by UAV during each simulation to test

5
0
5
4
2
0
2
0
1
2
3
X
Y
Route 2
Route 1
Route 3
5 4 3 2 1 0 1 2 3 4 5
5
4
3
2
1
0
1
2
3
X
(b) Top view (a) 3D view
5 4 3 2 1 0 1 2 3 4 5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
X
(c) Side view I
5 4 3 2 1
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Y
0 1 2 3
(d) Side view II

Fig. 11 Trajectories followed by chaotic control strategy for UAV.

Journal of Bionic Engineering (2009) Vol.6 No.3
278

the performance of the UAV under designed percep-
tion-action dynamics system. Fig. 12 and Fig. 13 show
the cumulative number of detected obstacles and the
volume explored by UAV in the four-buildings envi-
ronment (Fig. 11) respectively. The capability of
autonomous navigation of UAV is demonstrated from
the results.

0 1 2 3 4 5 6 7 8 9 10
0
5
10
15
20
25
30
Epochs 1000
N
u
m
b
e
r

o
f

d
e
t
e
c
t
e
d

o
b
s
t
a
c
l
e
s
Route 1
Route 2
Route 3

Fig. 12 Cumulative number of detected obstacles by UAV.

E
x
p
l
o
r
e
d

v
o
l
u
m
e

(
%
)

Fig. 13 Volume explored by UAV.

Subsequently, to evaluate the performance of the
proposed control scheme, a comparison with a tradi-
tional navigation control method is reported here. The
traditional navigation strategy chosen here is the Artifi-
cial Potential Field (APF) approach
[22]
. Also in this case
the UAV can use only local information, acquired from
its sensory system to react to the environment conditions
(i.e. PF). The parameters of the APF algorithm (e.g.
UAV speed, constraints for the movements) are chosen
in order to allow a comparison with our method. Route 1
is a good candidate as the trajectory for the comparison.
Fig. 14 and Fig. 15 show that the explored volume by
chaotic control method (CC) is generally bigger than
that by APF algorithm (APF) while the number of de-
tected obstacles by CC method is less. Anyway, the
results demonstrate that the chaotic control method
guarantees a higher exploration capability compared
with the traditional APF control method.

N
u
m
b
e
r

o
f

d
e
t
e
c
t
e
d

o
b
s
t
a
c
l
e
s

Fig. 14 Comparison of cumulative number of detected obstacles.

0 1 2 3 4 5 6 7 8 9 10
0
10
20
30
40
50
60
70
80
Epochs 1000
CC
APF

Fig. 15 Comparison of volume explored by UAV.

5 Conclusions
In this paper the problem of multi-sensory integra-
tion is treated using a new technique called chaotic bi-
onics control based on perception-action dynamics. This
approach takes inspiration from the Freemans bionic
theory of brain pattern formation, although it makes use
of a more abstract model, and is applied to the UAV
navigation control problem. The phenomenon of en-
coding information stabilizing the unstable orbits en-
dowed in a chaotic attractor is investigated. The 3D
multiscorll chaotic system was chosen for its simple
model and to extend the emerging multiscroll attractor to
3D varying also the number of scrolls. The feedback
from the environment is introduced by using a

Yu et al.: Autonomous Navigation for Unmanned Aerial Vehicles Based on Chaotic Bionics Theory
279
continuous multi-stimuli chaos control technique. Fi-
nally, simulation results demonstrate the capability of
autonomous navigation for UAV in complex spatial
environments.
Acknowledgements
This work was supported by the National High
Technology Research and Development Program of
China (863 Program) (2006AA12A108) Multi-sensor
Integrated Navigation in Aeronautics Field from the
Ministry of Science and Technology of China, CSC
International Scholarship (2008104769) from Chinese
Scholarship Council, as well as International Post-
graduate Research Scholarship Program
(2009800778591) from Australian Government.
References
[1] Ollero A, Merino L. Control and perception techniques for
aerial robotics. Annual Reviews in Control, 2004, 28,
167178.
[2] Dufresne S, Johnson C, Mavris D N. Variable fidelity con-
ceptual design environment for revolutionary unmanned
aerial vehicles. Journal of Aircraft, 2008, 45, 14051418.
[3] Samad T, Bay J S, Godbole D. Network-centric systems for
military operations in urban terrain: The role of UAVs. Pro-
ceeding of the IEEE, 2007, 95, 92107.
[4] Verschure P, Voegtlin T, Douglas R J. Environmentally
mediated synergy between perception and behavior in mo-
bile robots. Nature, 2003, 425, 620624.
[5] Verschure P, Althaus P. A real-world rational agent: Unifying
old and new AI. Cognitive Science, 2003, 27, 561590.
[6] Freeman W J. Simulation of chaotic EEG patterns with a
dynamic model of the olfactory system. Biological Cyber-
netics, 1987, 56, 139150.
[7] Freeman W J. The physiology of perception. Scientific
American, 1991, 264, 7885.
[8] Freeman W J. A neurobiological theory of meaning in per-
ception. Part I: Information and meaning in nonconvergent
and nonlocal brain dynamincs. International Journal of Bi-
furcation and Chaos, 2003, 13, 24932511.
[9] Freeman W J. How and why brains create meaning from
sensory information. International Journal of Bifurcation
and Chaos, 2004, 14, 515530.



[10] Li X, Li G, Wang L, Freeman W J. Study of a bionic pattern
classifier based on olfactory neural system. Journal of Bi-
onic Engineering, 2004, 1, 133140.
[11] Harter D, Kozma R. Chaotic neurodynamics for autonomous
agents. IEEE Transactions on Neural Networks, 2005, 16,
565579.
[12] Kozma R, Freeman W J. Basic principles of the KIV model
and its application to the navigation problem. Journal of
Integrative Neuroscience, 2003, 2, 125145.
[13] Bae Y, Kim J, Kim Y. Obstacle avoidance methods in the
chaotic mobile robot with integrated some chaos equation.
International Journal of Fuzzy Logic and Intelligent Systems,
2003, 3, 206214.
[14] Arena P, De Fiore S, Fortuna L, Patan L. Perception-action
map learning in controlled multiscroll systems applied to
robot navigation. Chaos: An Interdisciplinary Journal of
Nonlinear Science, 2008, 18, 116.
[15] Kozma R, Freeman W J. The KIV model of intentional
dynamics and decision making. Neural Networks, 2008, 22,
277285.
[16] Yalcin M E, Suykens J A K, Vandewalle J. True random bit
generation from a double scroll attractor. IEEE Transactions
on Circuits and Systems I: Regular Papers, 2004, 51,
13951404.
[17] Lu J, Chen G, Yu X, Leung H. Design and analysis of mul-
tiscroll chaotic attractors from saturated function series.
IEEE Transactions on Circuits and Systems I: Regular Pa-
pers, 2004, 51, 24762490.
[18] Lu J, Han F, Yu X, Chen G. Generating 3-D multi-scroll
chaotic attractors: A hysteresis series. Automatica, 2004, 40,
16771687.
[19] Ahmad W M. Generation and control of multi-scroll chaotic
attractors in fractional order systems. Chaos, Solitons and
Fractals, 2005, 25, 727735.
[20] Yalcin M E. Multi-scroll and hypercube attractors from a
general jerk circuit using Josephson junctions. Chaos, Soli-
tons and Fractals, 2007, 34, 16591666.
[21] Pyragas K. Continuous control of chaos by self-controlling
feedback. Physics Letters A, 1992, 170, 421428.
[22] Khatib O. Real-time obstacle avoidance for manipulators
and mobile robots. International Journal of Robotics Re-
search, 1986, 5, 9098.

You might also like