Effect of Random Parameter Switching on Commensurate

Fractional Order Chaotic Systems

Saptarshi Das1, Indranil Pan2, and Shantanu Das3

1. School of Electronics and Computer Science, University of Southampton, Southampton SO17 1BJ, United
Kingdom.
2. Department of Earth Science and Engineering, Imperial College London, Exhibition Road, SW7 2AZ, United
Kingdom.
3. Reactor Control Division, Bhabha Atomic Research Centre, Mumbai-400085, India.

Email: s.das@soton.ac.uk, saptarshi@pe.jusl.ac.in (S.Das*)

Corresponding authors phone number: +44(0)7448572598

Abstract
The paper explores the effect of random parameter switching in a fractional order (FO) unified chaotic
system which captures the dynamics of three popular classes of chaotic systems i.e. Lorenz, Lu and
Chens attractors. The appearance of chaos and hyper-chaos in such systems which continuously
switches from one family to the other has been investigated here for a commensurate fractional order
system scenario. Our simulation study shows that noise like random variation in the unified chaotic
system key parameter along with gradual decrease in the commensurate FO of the model is capable of
suppressing the chaotic fluctuations much earlier than the fixed parameter one. Such a noise-like
random switching mechanism could be useful for stabilization and control of chaotic oscillation in
many real-world applications.

Keywords:
Fractional order chaos; Lorenz-Lu-Chen system; Lyapunov exponent; noise-induced chaos; random
parameter switched chaos; unified chaotic system

1. Introduction
Fractional calculus has given impetus to the study of dynamical systems which give rise to
chaos and are applicable in diverse disciplines like physics, biology, economics etc. [1]. Fractional
order nonlinear systems governed by fractional order differential equations (FODEs) have been
shown to produce chaotic phenomena even for an overall order lesser than three which contradicted
with the classical theories of integer order (IO) nonlinear dynamical systems that chaos can only
manifest in a continuous time nonlinear dynamical system having an order at least three [2]. This is
because of the fact that each FO differ-integral operators are actually infinite dimensional systems
which manipulates the underlying dynamics of each state with a very high order linear filtering (under
a chosen rational approximation technique). However such investigations are not without their
pitfalls, as higher order rational approximations of such FO operators have also led to spurious
notions of chaos, termed as fake chaos [1]. Classical way of detecting chaos is by the study of the
Lyapunov exponents from the system structure or from the observation of the states in the case of
1
unavailability of systems governing equations [3]. Chaos can be detected when the system has at
least one positive Lyapunov exponent. If a system with four state equations has two positive
Lyapunov exponents the system is known as a hyper-chaotic system. After the exhaustive study of
fractional calculus in linear control theory for the last decade [4]-[5], it has pervaded into the
nonlinear dynamical systems theory as well [1]. For example many classical chaotic systems have
been extended with its analogous fractional order versions e.g. Chua system [2], Lorenz system [6],
Chen system [7], Lu system [8], unified chaotic system [9], Rossler system [10], Liu system [11],
Duffing oscillator [12], Van-der Pol oscillator [13], Lotka-Volterra model [14], financial system [15],
NewtonLeipnik system [16], Voltas system [17], Arnedo system [18], GenesioTesi system [19],
neuron network system [20], micro-electromechanical system (MEMS) [21], multi-scroll chaotic
attractors [22] and multi-wing chaotic attractors [23] etc. Similar traces of FO hyper-chaos have been
found in several systems like hyper-chaotic-Rossler system [10], Lorenz system [24], Chen system
[25] and Four-wing attractor [26]. Beside the extension of FO versions of many well-researched
chaotic systems, mentioned earlier, FODEs are commonly observed in various natural systems where
the temporal dynamics has got a long-memory behavior like 1/f noise, heavy-tailed distribution,
long-range dependency in the auto-correlation, power-law decay in excitation-relaxation systems (like
electrical circuits, spring-mass-damper system, chemical reactions etc.) instead of an exponential
envelope [4, 27]. Fractional dynamics is more evident in time domain representation of naturally
occurring oscillatory systems (like viscoelasticity, fluid flow etc.) where the notion of damping is
different with a decaying power-law envelope and Mittag-Leffler type oscillations [28, 29], instead of
sinusoidal oscillations within an exponentially decaying envelope.
The focus of the present study is to first observe chaos and hyper-chaos in the unified chaotic
system family that encompasses the dynamics of three classes of chaotic systems (i.e. Lorenz, Lu and
Chen) for different ranges of a single key system parameter (). Next, a random noise like fast
switching in the parameter is proposed, such that the systems behavior continuously jumps
between these three families of chaotic attractors. Simultaneously, the effect of decreasing the
commensurate fractional orders [1] in obtaining chaos and hyper-chaos have been elucidated to
discriminate the new phenomena in comparison with the traditional fixed parameter FO chaotic
systems, vastly reported in various literatures, mentioned above.
The rest of the paper is organized as follows. Section 2 briefly describes the unified FO
chaotic system and the simulation methodology. Section 3 shows the simulations for the random noise
like fast switching in the unified chaotic system parameter and finds the disappearance of chaos
particularly at low values of the commensurate FO. Similar explorations are done in section 4 for the
unified hyper-chaotic system. Section 5 sheds light on the Lyapunov exponent based analysis of the
existence of chaos in the proposed random parameter switched Unified system, followed by
discussion in section 6. The paper ends with the conclusion in section 7 and the references.

2. Fractional Order Random Parameter Unified Chaotic Systems

2.1. Unified Chaotic System

The integer order unified chaotic system (1) models the classical Lorenz-Lu-Chen family of
chaotic attractors with a common template and a single key parameter that determines which family
the chaotic system will belong to.

2
 x = ( 25 + 10 )( y x )
y = ( 28 35 ) x xz + ( 29 1 ) y
(1)
z = xy ( + 8 ) z 3

Here, [ 0,1] . The system behaves as Lorenz, Lu and Chen chaotic system for the following ranges of
parameter respectively.
[0, 0.8 ) Lorenz family
= 0.8 Lu system (2)
( 0.8,1] Chen family

Recent studies [9] have developed analogous fractional order chaotic system by replacing the integer
order derivative terms by equivalent fractional derivatives, as shown in (3) with the commensurate
fractional order ( 0,1] which still exhibits chaotic nature depending on the value of the order of the
FO nonlinear system .
d x
= ( 25 + 10 )( y x )
dt
d y (3)
= ( 28 35 ) x xz + ( 29 1 ) y
dt
d z
= xy ( + 8 ) z 3
dt

2.2. Basics of Fractional Calculus and FO Chaotic Systems

It has been shown in [9] and many other literatures that such FO systems exhibits chaos even
when the commensurate order is less than one i.e. the overall system order being lesser than three
which is impossible in classical integer order chaotic systems [2]. Numerically solving such FO
chaotic systems is a bit challenging as compared to the IO chaotic systems. The Adams-Bashforth-
Moulton predictor-corrector method is widely adopted in recent literatures to avoid appearance of
fake chaos or loss of chaos due to truncation or rational approximation by frequency domain methods
of realizing fractional differ-integrals. But for noisy FO chaotic systems and fast random switched
parameter systems, extension of such time domain numerical solvers similar to Adams-Bashforth-
Moulton method has not been explored yet. Petras [1] has suggested that as an alternative way, the FO
differential equations may be converted to a set of integral equations with fractional derivative less
than unity as shown in equation (4) for efficiently solving such equations. In this circumstances, the
fractional derivatives can be rationalized using Oustaloups/modified Oustaloups or any other
frequency domain approximation methods [5].

x ( t ) = 0 D t1 ( 25 + 10 )( y x ) dt
t

0
(4)
y ( t ) = 0 D t1 ( 28 35 ) x xz + ( 29 1 ) y dt
t

z ( t ) = 0 D t1 xy ( + 8 ) z 3 dt
t
0

1
where, 0 Dt is Riemann-Liouville fractional differentiation operator.

In the present study, we have used the modified Oustaloups recursive approximation for each
fractional derivative with a 30th order approximation within a chosen frequency range of
[ l , h ] = 10 4 ,10 4 H z using the FOMCON Toolbox [30] in Matlab/Simulink environment to simulate
such FO chaotic systems as also suggested by Petras [1]. The modified Oustaloups approximation

3
converts the fractional differ-integrals in Riemann-Liouville definition (5) in equivalent frequency
domain Infinite Impulse Response (IIR) filters given by (6).
1 dm f ( )
D t f ( t ) =
t
( t ) d , ( m 1) < < m , +
( m ) dt m m +1 (5)
a
a

L { D x ( t )} = s X ( s ) , x ( 0 ) = 0, x 1 ( 0 ) = 0,

Here, s and t denotes the Laplace and time domains respectively and the initial conditions in the
Laplace transform of fractional derivatives are assumed to be zero [5].

d f h d f s2 + b f hs
s G p
d f (1 ) s + b f h s + d f
b 2
f (6)
+2k 2k
N
s + k b f h 2 N +1 d f b 2 N +1
Gp = s + k
,k =
d , =
b
k = N f f

The choice of refined Oustaloups filter parameters (bf and df) in (6) has been described in
[31] in an optimization based framework where it has been shown that the constant phase
performance remains almost consistent with a fixed b f d f ratio. The particular advantage of using
modified Oustaloups approximation instead of original Oustaloups approximation (implemented in
the other Matlab based Toobox for FO systems, called Ninteger [32]) is that the former gives less
ripple in the frequency domain, maintains a constant phase for a wider frequency range and gives
good approximation even near the chosen lower/upper frequency boundaries [31], [5]. For the present
simulation study, all the nonlinear integer/fractional order ordinary differential equations (ODEs) are
integrated with the fourth order Runge-Kutta method with fixed step size of 0.001 sec.

Fig. 1. Phase portraits of integer order ( = 1) and commensurate fractional order ( = 0.9) chaotic
Lorenz system ( = 0.7)

As mentioned earlier, the unified chaotic system behaves as the family of Lorenz, Lu and
Chen systems for the parameter ranges given in (2). The phase portraits of the three integer order
4
chaotic systems (Lorenz-Lu-Chen) and that of their commensurate fractional order counterparts (with
= 0.9) are shown in Figs. 1-3 respectively with an initial condition of ( x0 , y0 , z0 ) = (1,3, 2 ) . It is found
that the unified chaotic system in all its three subclasses exhibits chaotic behaviour in the phase space
while lowering the commensurate FO till 0.85 (Fig. 4). For further decrease in commensurate FO, the
phase space trajectory stabilizes in one of the stable equilibrium points as shown in Fig. 4 with =
0.8. Therefore, for FO systems, chaos can appear even in a system having an overall order
0.85 3 = 2.55 . Tavazoei and Haeri in [33] have shown the bounds of observing chaos in
commensurate and incommensurate FO unified chaotic system and in general, the gradual decrease in
FO causes disappearance of chaos.

Fig. 2. Phase portraits of integer order ( = 1) and commensurate fractional order ( = 0.9) chaotic Lu
system ( = 0.8)

5
Fig. 3. Phase portraits of integer order ( = 1) and commensurate fractional order ( = 0.9) chaotic
Chen system ( = 0.9)

Fig. 4. Disappearance of chaos in FO unified chaotic system for commensurate order 0.8

3. Random Noise Like Fast Switching in the Unified Chaotic System Parameter
Here, we explore a new class of chaotic systems with random switching in its one parameter.
The IO and FO unified chaotic system is chosen for the simulation study with a consideration of

6
random switching in parameter U ( 0,1) . Therefore, the key system parameter of the unified
chaotic system is randomly drawn from a uniform probability density function (pdf) such that at each
time instant it behaves in a particular way but at the next time instant it jumps to another family of
chaotic systems according to equation (2). The small wiggles in the phase space trajectories even for
the IO case is evident in Fig. 5 due to random variation in the unified chaotic system parameter , as
opposed to the smooth phase space trajectories in IO unified chaotic systems in Figs 1-3. The
aberrations in the phase space trajectories are more evident for lower value of the commensurate
fractional order i.e. for decreasing the value of from one to 0.95 and then 0.9 in Fig. 5-7. The change
in the phase space trajectories for such random noise like fast parameter switching FO systems
indicates towards a new type of chaotic behaviour, as compared to that shown in Fig. 1-3, for classical
fixed parameter IO/FO chaotic systems. Systems exhibiting this kind of behaviour are henceforth
termed as IO/FO random parameter switched chaotic systems in the rest of this paper.

Fig. 5. Phase portrait of integer order ( = 1) unified chaotic system with random variation in with
wiggly trajectories

7
Fig. 6. Phase portrait of fractional order unified chaotic system with random variation in with =
0.95 with wiggly trajectories

It is also expected that for such new class of random parameter FO chaotic systems the
obtained state-time series would be totally different compared to the classical versions. Here, starting
from the same initial condition ( x0 , y0 , z0 ) and same initial parameter 0 = 0.8 , the systems state
evolves in a different way and diverges very quickly from each other due to the presence of random
switching in the parameters of the chaotic system. The three state trajectories under this situation have
been compared in Fig. 8, as an illustrative example. Although the deterministic long term prediction
property holds for the fixed parameter system since its underlying model is known but for the latter
case due to the uncertainty introduced in the value of key system parameter () due to noisy switching
at each time instant, long term prediction becomes difficult. As can be seen from the phase portraits in
Figs. 5-7, their shapes qualitatively match the traditional cases with constant parameters in Figs. 1-3.
However, a look at the individual time series in Fig. 8 confirms that the evolution of the state
trajectories are entirely different from the traditional case.

8
Fig. 7. Phase portrait of fractional order unified chaotic system with random variation in with =
0.9

Another interesting fact can be observed for such switched parameter chaotic systems. While
gradually decreasing the commensurate FO, the chaos disappears much earlier than the fixed
parameter FO chaotic systems. This typical phenomenon has been shown in Fig. 9 where the phase
space trajectories stabilizes in one of the equilibrium points for a low commensurate FO of = 0.85,
whereas the fixed-parameter FO chaotic systems (Lorenz, Lu and Chen sub-classes) in Fig. 4 still
exhibit chaos at that low fractional order of = 0.85.

9
Fig. 8. Comparison of state time series for random parameter and fixed parameter FO unified chaotic
system

Fig. 9. Early disappearance of chaos at = 0.85 due to random parameter switching in commensurate
FO unified chaotic system

10
 4. Fractional Order Random Parameter Unified Hyper-chaotic Systems

4.1. Hyper-chaotic Fractional Order Fixed Parameter (Lu) System

Analogous hyper-chaotic system for the unified chaotic system (3) is described by four
commensurate FO state equations. The overall system is defined by (7) with an additional equation
involving state w [34] which is originally derived from Chen system.
d x
= ( 25 + 10 )( y x )
dt
d y
= ( 28 35 ) x xz + ( 29 1 ) y (7)
dt
d z
= xy ( + 8 ) z 3
dt
dw
= x + 0.5
dt

Fig. 10. Phase portraits for IO (=1) Hyper-chaotic Lu system

Simulation studies are reported with an initial condition of ( x0 , y0 , z 0 , w0 ) = ( 4, 5, 6, 3 ) for

exploring the effect of FO dynamics along with random parameter switching in the unified hyper-
chaotic system. Fig. 10 shows the phase portraits of the IO classical hyper-chaotic system in 2D and
3D phase space diagrams showing highly chaotic motion. As a next step of exploration, the
commensurate FO of the hyper-chaotic system (7) is gradually decreased from 0.9 to 0.7. It is
observed that there is a sudden disappearance of chaos at 0.8 for this system also, as explored in
the earlier section with FO unified chaotic systems. The disappearance of chaos due to decrease in the
overall system order is evident since the phase portraits becomes more regular suggesting regular
oscillation or damping of the oscillations which is explored in Fig. 11-13 for the hyper-chaotic Lu
system ( = 0.8), as a representative candidate of the unified hyper-chaotic system family (7).
Although the phase portraits clearly show damping of chaotic oscillation, the disappearance of chaos

11
has to be alternatively verified by looking at the largest Lyapunov exponent (LLE), as reported in the
next section.

Fig. 11. Phase portraits for Hyper-chaotic commensurate FO (=0.9) Lu system (=0.8)

Fig. 12. Phase portraits for Hyper-chaotic commensurate FO (=0.8) Lu system (=0.8)

12
Fig. 13. Phase portraits for Hyper-chaotic commensurate FO (=0.7) Lu system (=0.8)

4.2. Random Parameter Fractional Order Unified Hyper-chaotic System

Next, the key system parameter () of unified chaotic system is randomly switched by
sampling it from a uniform pdf as done previously for the unified chaotic system. The IO random
parameter chaotic system in Fig. 14 although shows chaos like motion but the deterministic nature of
the system is lost due to random parameter switching and the model therefore cannot be used for
making long term prediction by only knowing the initial state x0 . The corresponding phase space
diagrams along with gradual decrease in commensurate FO from 0.9 to 0.6 are shown in Fig. 15-18.
In this case also, the chaos disappears below 0.8, making the phase space trajectories quite well-
behaved and regular. In these cases, the lower the value of , much smoother phase portraits are
observed which finally settles down to a stable equilibrium point.
Due to random parameter switching, especially in low value of commensurate FO, the phase
space trajectories show some noisy or wiggly behaviour while approaching towards the stable
equilibrium points. Therefore, it is important to find out the particular effect of introducing random
parameter switching compared to the classical fixed parameter FO unified hyper-chaotic system.
Also, it is important to find out at which commensurate FO order, the chaos first disappear for random
parameter system while the fixed parameter FO counterpart still exhibits chaotic motion. For the sake
of comparison, the hyper-chaotic FO Lu system ( = 0.8) is explored among the unified hyper-chaotic
system family and its phase space characteristics is compared with the random parameter hyper-
chaotic system with same initial conditions, in Fig. 19-20.

13
Fig. 14. Hyper-chaotic unified IO system ( = 1) with random parameter variation

Fig. 15. Hyper-chaotic unified system ( = 0.9) with random parameter variation

Here, it has been shown that the random parameter switching in fractional order hyper-chaotic
system destabilizes the chaos, even at a higher value of commensurate fractional order ( = 0.83) than
the fixed parameter FO chaotic Lu ( = 0.8) systems with the same value of . The regularity in the
phase space diagrams at = 0.83 for the random parameter unified chaotic system is evident although
the fixed parameter counterparts still shows chaotic motion in Fig. 20. This shows that such random
parameter switching may be useful for control or suppression of chaotic phenomena in many naturally
occurring chaotic/hyper-chaotic higher dimensional systems [35], where the task is to suppress
chaotic oscillations or neutralize chaotic oscillations by noise [35-36]. Previously, a similar concept
has been introduced in [37]-[40] known as the noise induced chaos where the chaotic behaviour can
14
be an effect of random parameter variation in a nonlinear system. The present approach is a bit
different since the random variation in the key parameters () makes the system arbitrarily switch
between the three families viz. Lorenz-Lu-Chen systems. The presence and disappearance of chaotic
motion in the obtained time series in the previously explored fixed and random parameter system are
next verified using the numerical computation of Lyapunov exponents.

Fig. 16. Hyper-chaotic unified system ( = 0.8) with random parameter variation

Fig. 17. Hyper-chaotic unified system ( = 0.7) with random parameter variation

15
Fig. 18. Hyper-chaotic unified system ( = 0.6) with random parameter variation

Fig. 19. Comparison of fixed/random parameter FO unified hyper-chaotic system for = 0.85

5. Lyapunov Exponents of the Time-Series of Random Parameter Switched

Chaotic Systems
Investigation of chaos is mostly done using its genuine signature known as the Lyapunov
exponent. There has been significant amount of research in past to calculate the largest Lyapunov
exponent (LLE) from a finite length time-series which is capable of discriminating the underlying

16
dynamical behavior of a data due to random noise and chaotic motion. Lyapunov exponent greater
than zero indicates the presence of chaotic fluctuation in a measured dataset. In most realistic cases, a
measured time series is corrupted with white or colored noise and there could be spurious results for
estimating the LLE depending on the strength of the noise added to the system. Investigation of the
additive noise with chaotic time series with different signal to noise ratio (SNR) levels has been
reported in Rosenstein et al. [41].

Fig. 20. Comparison of fixed/random parameter FO unified hyper-chaotic system for = 0.83

In the present paper, the situation is different since the dynamics of chaotic time series
masked by additive measurement noise (with different noise strength or variance) is not the main
focus of the investigation here unlike the research reported in Rosenstein et al. [41]. But here,
dynamics of the chaotic attractor is determined by random fluctuation in the key system parameters
(). As discussed above, the system has a similar framework like a noise induced chaotic system [37]-
[40]. Therefore the exploration reported here is markedly different from that reported with additive
noise [36] in various known family of chaotic systems. In addition, the effect of fractional dynamics
on such random parameter switching in the chaotic systems and their corresponding Lyapunov
exponents have also been investigated here.
From a single time-series representation of the measured state xi , with the help of time delay
embedding, the attractor dynamics is normally reconstructed. If x ( ) be the evolution of the state from
some initial condition x ( 0 ) , then the largest Lyapunov exponent is calculated as (8).

1 x ( ) x ( )
= lim ln , x ( 0 ) x ( 0 ) = (8)

0
Now for varying relative time ( ) the Lyapunov exponents are calculated as (9).
17
 1 N 1
S ( ) = ln
N i =1 U i
dist ( x x , )
i, j (9)
x j Ui

where, Ui is the neighborhood of xi with diameter and the dist ( xi , x j , ) is the distance between a
trajectory xi and a neighbor x j , after the relative time . The plot of vs. S ( ) helps to determine
whether the embedding dimension is selected in a proper way if the curve is more or less horizontal or
smooth [42]. Gao et al. [43] proposed a technique using scale-dependent Lyapunov exponent to
distinguish between chaos, noisy chaos and noise induced chaos where the presence of the chaos
could be confirmed by observing a plateau over multiple scales. The present work takes a similar
approach where the Lyapunov exponents with varying relative time ( ) is reported as also reported in
[42]. It is a well discussed topic to distinguish chaos from noise using Lyapunov exponents and there
are significant amount of research evidences in the past [44]-[47]. Kinser [45] reported that the LLE
of noise may be a very large number and other nonlinear measures may be needed. Therefore in the
present scenario, the LLE, Lyapunov exponents for varying relative time S ( ) and Shannon entropy
analysis both have been reported together, in order to understand the nature of the time series obtained
from these new classes of dynamical systems.

Fig. 21. Lyapunov exponents for fixed parameter chaotic systems.

Apart from the LLE, another measure widely used in the investigation of randomness of the
information and nonlinearity is the Shannon entropy. It can be easily calculated from the wavelet
decomposition of the time-series x ( t ) in several orthonormal basis ( xi ) such that
E ( x ) = E ( xi ) = xi2 log ( xi2 ) and with the convention of E ( 0 ) = 0 log ( 0 ) = 0 . Here, the Lyapunov
i i

exponents and wavelet entropy were calculated using the 12500 samples of the sampled sequences of
the first state variable (x) using the delay embedding technique as in [41].

18
 Due to the difficulty of overflowing memory while calculating the LLE using large number of
data samples in the state time series, the signal is down-sampled by a factor of four, for various cases
of the family of systems under investigation as reported in Table 1. In order to obtain a fair
comparison, the LLE and entropy for a white Gaussian noise and chaotic Logistic map are also
calculated as two extreme examples of pure noise and pure chaos respectively, without mixture so that
later the deterministic and stochastic dynamics (along with their overlap) on the noise induced chaos
in commensurate fractional nonlinear dynamical systems can be quantified. The calculated LLE and
entropy show a clear signature of chaos for the Logistic map due to significantly large and positive
LLE while the high randomness is affirmed from the large entropy value for white Gaussian noise.
For all the three subclasses of the unified chaotic systems i.e. fixed parameter FO Lorenz-Lu-Chen,
the LLE decreases and entropy increases with gradual decrease in the commensurate fractional order
as shown in Table 1. The disappearance of chaos is evident from sudden drop in the LLE at = 0.8
for the fixed parameters FO unified system. A similar behavior can be observed for the random
parameter case also where the chaos disappears much faster at = 0.85. These time series based
estimates confirm the graphical findings in section 2 and section 3 as well. For the fixed parameter
hyperchaotic FO Lu systems, there is a sudden drop in LLE at commensurate FO, = 0.8 and the
corresponding LLE values for the random parameter case is relatively lower than the corresponding
fixed parameter cases. In principle for stable nonlinear systems i.e. damping of chaotic oscillations
should exhibit a negative LLE but in most cases when the chaos stabilizes, it still show a positive LLE
which should not be confused with weak chaotic behavior. Since, the estimation of LLE is attempted
from the finite number of samples of the state time series and not the system structure itself (using
Jacobian), the involved numerical method may not often give exactly zero or negative value but gives
a LLE which is close to zero and also a sudden drop in the Lyapunov exponents for varying relative
time. The main motivation behind applying a single state time series based LLE calculation instead of
the conventional structure based LLE estimation method is that under varying structure or random
switching between different differential equation structures the latter technique does not hold [48, 49].
Therefore, using the observed time series of the state variables via the time delay embedding is a
viable alternative to compute the LLE for this new interesting FO semi-deterministic chaotic system.
The delay embedding techniques are mostly used on experimental data when the underlying data
generation process or the governing differential equations are not precisely known or there is noise in
the system. Although it is not very popular for simulation studies on deterministic chaos, as the
structure based LLE computation is preferred without the presence of noisy fluctuations.
Next the Lyapunov exponents for varying relative time - S ( ) is explored [42] using equation
(9), for each of the systems in Figs. 21-24 while the relative time has been increased from = 1 20 .
Fig. 21 shows that the Lyapunov exponents are having significantly positive value for = 1 and =
0.9 for all the three cases of Lorenz-Lu and Chen system. But with = 0.8, the estimated Lyapunov
exponents are much smaller and sometimes become negative, especially at higher value of . Fig. 22
explores the case of discrete time iterators like white noise (with zero mean and standard deviation
= 100 ) or chaotic Logistic map ( xn+1 = rxn (1 xn ) , r = 4, x0 = 0.3). It is evident from Fig. 22 that the
Lyapunov exponents for the white noise is very small and often becomes negative for lower values of
relative time ( = 2 10 ) whereas the Logistic map shows a large positive Lyapunov exponent for
= 1 7 . This comparison helps to get an intuitive analysis of noisy sequence, deterministic chaotic
sequence as well as the semi-deterministic ones, as in the present case. The other Lyapunov exponents
in Fig. 22 show that the hyper-chaotic fixed parameter Lu system ( = 0.8 ) exhibits chaos with = 1
and = 0.9 whereas the values suddenly drop at = 0.8 and = 0.7 indicating disappearance of
chaos. In Fig. 23, the Lyapunov exponents of the random parameter unified system show that

19
although there is a strong chaotic behaviour for = {1, 0.95, 0.9}, but at = 0.85 the chaotic
behaviour is lost due to drop in the Lyapunov exponent values which confirms the studies reported in
section 3. A similar case is observed at = 0.8 and below in Fig. 24 for the random parameter hyper-
chaotic system confirming the findings reported in section 4. The S ( ) plots along with the LLEs in
Table 1 confirm the presence or disappearance of chaotic behaviour of each FO system.

Table 1: Estimation of the LLE and Shannon Entropy for the fixed/random parameter chaotic/hyper-
chaotic systems
Shannon
System LLE Entropy

White Gaussian noise ( N ( 0, 2 ) , = 100) 0.00092741 2.05103

Logistic map ( xn+1 = rxn (1 xn ) , r = 4, x0 = 0.3) 0.6924 -1.21109

IO chaotic Lorenz system ( = 1) 0.1122 -4.26106
FO chaotic Lorenz system ( = 0.9) 0.1105 -3.63106
FO chaotic Lorenz system ( = 0.8) 0.0115 -3.52106
IO chaotic Lu system ( = 1) 0.1142 -4.43106
FO chaotic Lu system ( = 0.9) 0.1165 -3.62106
FO chaotic Lu system ( = 0.8) 0.0089 -3.46106
IO chaotic Chen system ( = 1) 0.1157 -4.42106
FO chaotic Chen system ( = 0.9) 0.1188 -3.59106
FO chaotic Chen system ( = 0.8) 0.0157 -3.39106
Random parameter IO unified chaotic system ( = 1) 0.1078 -4.16106
Random parameter FO unified chaotic system ( = 0.95) 0.1046 -3.88106
Random parameter FO unified chaotic system ( = 0.9) 0.1071 -3.65106
Random parameter FO unified chaotic system ( = 0.85) 0.0779 -3.64106
IO hyper-chaotic Lu system ( = 1) 0.1139 -4.42106
FO hyper-chaotic Lu system ( = 0.9) 0.1107 -3.66106
FO hyper-chaotic Lu system ( = 0.8) 0.0485 -3.27106
FO hyper-chaotic Lu system ( = 0.7) 0.0072 -3.06106
Random parameter IO unified hyper-chaotic system ( = 1) 0.1084 -4.26106
Random parameter FO unified hyper-chaotic system ( = 0.9) 0.0801 -3.62106
Random parameter FO unified hyper-chaotic system ( = 0.8) 0.0341 -3.43106
Random parameter FO unified hyper-chaotic system ( = 0.7) 0.0375 -3.24106
Random parameter FO unified hyper-chaotic system ( = 0.6) 0.0661 -3.14106
20
Fig. 22. Lyapunov exponents for fixed parameter hyper-chaotic systems and discrete time iterators
(white noise and Logistic map).

Fig. 23. Lyapunov exponents for random parameter chaotic systems.

Here, Table 1 is especially useful in understanding how the quantitative measure of chaotic
behaviour and randomness change with the decrease in the commensurate FO and between similar
fixed/random parameter systems. Table 1 also explains the trade-off between the deterministic and
stochastic dynamics under the same commensurate order for the fixed and random parameter
21
systems. For a specific choice of , it is observed that the LLE becomes lower (indicating a weaker
chaos) and entropy becomes higher (stronger stochastic dynamics or influence of noise) for the
random parameter systems than the corresponding fixed parameter systems.

Fig. 24. Lyapunov exponents for random parameter hyper-chaotic systems.

6. Discussion

6.1. Novelty of the Present Approach over Existing Theory of Noisy Chaos

There have been already many studies on discriminating noisy chaos, noise induced chaos and
fractional order stochastic processes (1/f) or higher order colored noise in previous literatures like
[43, 44]. It has also been shown in [36], that adding white or correlated noise makes a chaotic time
series impossible to be identified as chaos by standard LLE method which is known as titration of
chaos by noise. In such cases, the regular smooth structure of the attractor dynamics in phase space
breaks down and it no longer shows chaotic oscillations. Here we explore a different phenomenon of
noise induced chaos in FO unified system when the key parameter () was switched randomly along
with gradual decrease in the commensurate FO of the system. In the present study, at each time instant
the key parameter was sampled from a uniform distribution that determines, which family within the
uniform chaotic attractor (2), the system will it belong to at that particular time instant.
Also, there are other methods to investigate deterministic chaos e.g. the 0-1 test, apart from
the popular method of LLE. But the 0-1 test of chaotic dynamics was originally developed to show
chaotic dynamics with large number of noise free data [50]. Investigations by Gottwald and
Melbourne [51] suggests that the 0-1 method can still detect the underlying deterministic chaotic
dynamics if the noise-level is sufficiently small (10%) and not correlated with the systems dynamics.
In our present study, the main idea is not to investigate weak chaos masked by additive noise but the
when the whole systems dynamics is governed by noise like parameter switching. In such cases of
22
noise induced FO chaos, the 0-1 test is not guaranteed to detect the underlying chaotic behavior.
Therefore we restrict our study to determine the domination of the deterministic and stochastic
dynamics using the two popular methods Lyapunov exponents and entropy criteria only. Also,
further investigation is needed to understand why the chaos disappears with decrease in the
commensurate FO () of the system. Almost in all FO chaotic dynamical systems, the chaos is found
for a certain range of parameters and not under all parameter settings. Previous investigations of FO
chaos in [6]-[26] mostly reported similar approach of the existence of chaos by simulation study,
under various parameter settings. The motivation of the present paper is to investigate the range of
system parameters for which the chaos persists in a uniformly distributed noise induced FO unified
chaotic systems, although the physical interpretation of disappearance of chaos especially at low is
still an open question. Also, the distribution of the key parameter () is chosen from a uniform pdf
rather than a Gaussian or other pdfs in order to sample almost equal number of times from all parts of
the pdf and also to strictly enforce that the key parameter lies within zero to one, signifying switching
amongst Lorenz, Chen and Lu families as per equation (2).
Also, the different dynamical characteristics, in terms of sustained or damping of the chaotic
excursions of the states, have been observed in the present study while using different values of the
commensurate FO. This raises a question whether the suppression of chaos is a generic characteristic
for large scale nonlinear dynamical systems in the presence of noise and random parameter switching
or not. Molgedey et al. [35] have shown that noise has different effects on the chaotic dynamics
depending on the dimensionality of the system. For low dimensional systems, noise favors chaos,
while in high dimensional systems noise impairs the flow of information and inhibits chaos. In this
paper, the consideration of the fractional dynamics in the state equations makes the system infinite
dimensional (which is approximated by a very high dimensional constant phase filtering). The aim of
the paper is to investigate whether the suppression of chaos due to noise is a generic phenomenon
among all the members of the family of FO attractors (with different values of ). Previous literature
like [35], [36] addresses the stabilizability and detectability of chaos in the presence of varying
degrees of additive noise which is different from the random parameter switching within the three
families of attractors as done in this paper. The reported simulation examples can be viewed as a first
proof of concept study for this specific type of phenomena and a more in depth study using analytical
techniques like switched and hybrid dynamical systems are required in the future to lay the foundation
of the underlying mathematical basis, which may be pursued as a scope for future work. One of the
limitations in applying the switched chaotic dynamical systems theory [52], in the present scenario, is
that the Lipschitz continuity must be maintained at each of the switching instants [53]. In other words
there cannot be any kind of arbitrarily fast switching for the stabilization results to hold. There is also
a consequent concept of dwell time which puts an analytical bound on the rate of switching dynamics
[53]. In this paper, a noise like switching in is at every sampling instant of the numerical integration
in (4), unlike specific time instances reported in [52] which employ a zero order hold mechanism to
satisfy the conditions of Lipschitz continuity. This might pose problems for application of the
established analytical techniques in the present random parameter switched chaotic systems of integer
[52] and fractional order. Future work might look at appropriate mathematical refinements to address
this issue of stability analysis in the presence of fast switching.

6.2. Physical Significance of the Observation

It is well known that experimentally observed time domain trajectories of many real world
processes as found in diverse disciplines (like finance, biology) are often characterized by noise like
fluctuations. Therefore modelling the dynamical characteristics through ordinary (integer or fractional

23
order) nonlinear differential equations inherently use some kind of approximations due to the
smoothness of the differential equation solutions. Most of the literature tries to capture the noisy
phenomena by incorporating additive or multiplicative random variable in the state variables of the
differential equations. The present work shows how the behavior of the FO dynamical system changes
under the influence of random parameter vis-a-vis its constant counterpart. It has been shown in the
previous sections that randomly modulating the key parameter () results in the disappearance of
chaos. This is an important observation and may have many future applications, especially with
respect to chaos control and synchronization. Recently a special kind of random parameter chaotic
system, known as switched chaotic system has been explored in [52]. A single global controller has
also been designed analytically based on Lyapunov stability theory for the synchronization of chaos in
a wide variety of scenarios. The presented results adds value to the existing knowledge of such
random parameter chaotic systems in integer order [52] since there might not be any need to design
any external controller at all using the present design. It would be possible to selectively modulate
(by simply introducing randomness in) some key parameters of a chaotic system to stabilize it.
From mathematical perspective, the results in this paper fuses three major branches of
research in differential equations nonlinear chaotic differential equations, fractional order
differential equations and stochastic differential equations. This can serve as a building block for a
new type of design paradigm which leverages the strength of these three distinct disciplines in
differential equation based modelling of natural processes viz. sensitivity to initial condition (as
modelled by chaotic nonlinear differential equations), long memory process and rich higher order
dynamics (as modelled by fractional order differential equations) and stochastic fluctuation in the
state variables (as modelled by stochastic differential equations).

7. Conclusion
A new class of commensurate FO systems has been explored in this paper known as random
parameter switched chaotic or hyper-chaotic systems. The effect of random parameter switching in
the key parameter () of unified system or the noise like variation in also induces chaotic fluctuation
under different commensurate FO system have been studied. Simulation results and illustrations for
the unified chaotic or hyper-chaotic system show that chaotic behavior disappears much early (at a
higher value of commensurate FO) for random parameter switched chaotic systems while gradually
decreasing the commensurate order. The corresponding fixed parameter case still shows chaotic
behavior in this regime. The introduction of noise induced parameter switching in the chaotic system
and its effect in the titration of chaos [36], in other class of chaotic systems may be explored in future
studies. Here, we also show that the random switching of the key parameter () of the FO unified
chaotic system could be used to damp out the oscillations. Further engineering applications of the
proposed system could be directed towards stabilization or control of chaos using the theory of
switched dynamical systems. In addition, the noisy sequence for , along with its rate of switching
and the commensurate FO of the system could be used in various future applications in secure
communication, by serving as additional parameters in the design of the security keys.

References
[1] I. Petras, Fractional-order nonlinear systems: modeling, analysis and simulation, Springer-
Verlag, Berlin, 2011.
[2] T.T. Hartley, C.F. Lorezo, and H.K. Qammer, Chaos in a fractional order Chuas system,
Circuits and Systems I: Fundamental Theory and Applications, IEEE Transactions on, vol. 42, no. 8,
pp. 485-490, Aug. 1995.
24
[3] A. Wolf, J.B. Swift, H.L. Swinney, and J.A. Vastano, Determining Lyapunov exponents from a
time series, Physica D: Nonlinear Phenomena, vol. 16, no. 3, pp. 285-317, Jul. 1985.
[4] S. Das, Functional fractional calculus, Springer-Verlag, Berlin, 2011.
[5] C.A. Monje, Y.Q. Chen, B.M. Vinagre, D. Xue, and V. Feliu, Fractional-order systems and
control: fundamentals and applications, Springer-Verlag, London, 2010.
[6] I. Grigorenko and E. Grigorenko, Chaotic dynamics of the fractional Lorenz system, Physical
Review Letters, vol. 91, no. 3, pp. 034101, Jul. 2003.
[7] C. Li and G. Chen, Chaos in the fractional order Chen system and its control, Chaos, Solitons &
Fractals, vol. 22, no. 3, pp. 549-554, Nov. 2004.
[8] J.G. Lu, Chaotic dynamics of the fractional-order L system and its synchronization, Physics
Letters A, vol. 354, no. 4, pp. 305-311, Jun. 2006.
[9] X. Wu, J. Li, and G. Chen, Chaos in the fractional order unified system and its synchronization,
Journal of the Franklin Institute, vol. 345, no. 4, pp. 392-401, Jul. 2008.
[10] C. Li and G. Chen, Chaos and hyperchaos in the fractional-order Rssler equations, Physica A:
Statistical Mechanics and its Applications, vol. 341, pp. 55-61, Oct. 2004.
[11] X.-Y. Wang and M.-J. Wang, Dynamic analysis of the fractional-order Liu system and its
synchronisation, Chaos, vol. 17, pp. 033106, 2007.
[12] X. Gao and J. Yu, Chaos in the fractional order periodically forced complex Duffings
oscillators, Chaos, Solitons& Fractals, vol. 24, no. 4, pp. 1097-1104, May 2005.
[13] R. Barbosa, J.A.T. Machado, B.M. Vinagre, A.J. Calderon, Analysis of the Van der Pol
oscillator containing derivatives of fractional order, Journal of Vibration and Control, vol. 13, no. 9-
10, pp. 1291-1301, Sep. 2007.
[14] S. Das and P.K. Gupta, A mathematical model on fractional LotkaVolterra equations, Journal
of Theoretical Biology, vol. 277, no. 1, pp. 1-6, May 2011.
[15] W.-C. Chen, Nonlinear dynamics and chaos in a fractional-order financial system, Chaos,
Solitons& Fractals, vol. 36, no. 5, pp. 1305-1314, Jun. 2008.
[16] L.-J.Sheu, H.-K. Chen, J.-H. Chen, L.-M. Tam, W.-C. Chen, K.-T. Lin, and Y. Kang, Chaos in
the NewtonLeipnik system with fractional order, Chaos, Solitons& Fractals, vol. 36, no. 1, pp. 98-
103, Apr. 2008.
[17] I. Petras, Chaos in the fractional-order Voltas system: modeling and simulation, Nonlinear
Dynamics, vol. 57, no. 1-2, pp. 157-170, Jul. 2009.
[18] J.G. Lu, Chaotic dynamics and synchronization of fractional-order Arneodos systems, Chaos,
Solitons & Fractals, vol. 26, no. 4, pp. 1125-1133, Nov. 2005.
[19] J.-G. Lu, Chaotic dynamics and synchronization of fractional-order GenesioTesi systems,
Chinese Physics, vol. 14, no. 8, pp. 1517, Aug. 2005.
[20] S. Zhou, H. Li, and Z. Zhu, Chaos control and synchronization in a fractional neuron network
system, Chaos, Solitons& Fractals, vol. 36, no. 4, pp. 973-984, May 2008.
[21] A.M. Tusset, J. M. Balthazar, D. G. Bassinello, B. R. Pontes Jr, and Jorge Luis Palacios Felix,
Statements on chaos control designs, including a fractional order dynamical system, applied to a
MEMS comb-drive actuator, Nonlinear Dynamics, vol. 69, no. 4, pp. 1837-1857, Sep. 2012.
[22] W.M. Ahmad, Generation and control of multi-scroll chaotic attractors in fractional order
systems, Chaos, Solitons& Fractals, vol. 25, no. 3, pp. 727-735, Aug. 2005.
[23] C. Zhang and S. Yu, Generation of multi-wing chaotic attractor in fractional order system,
Chaos, Solitons & Fractals, vol. 44, no. 10, pp. 845-850, Oct. 2011.
[24] X.-Y. Wang and J.-M. Song, Synchronization of the fractional order hyperchaos Lorenz systems
with activation feedback control, Communications in Nonlinear Science and Numerical Simulation,
vol. 14, no. 8, pp. 3351-3357, Aug. 2009.

25
[25] X. Wu and Y. Lu, Generalized projective synchronization of the fractional-order Chen
hyperchaotic system, Nonlinear Dynamics, vol. 57, no. 1-2, pp. 25-35, Jul. 2009.
[26] S. Dadras, H.R. Momeni, G. Qi, and Z.L. Wang, Four-wing hyperchaotic attractor generated
from a new 4D system with one equilibrium and its fractional-order form, Nonlinear Dynamics, vol.
67, no. 2, pp. 1161-1173, Jan. 2012.
[27] S. Das, and I. Pan, Fractional order signal processing: introductory concepts and applications,
Springer, Heidelberg, 2012.
[28] S. Das, I. Pan, K. Sur, and S. Das, Artificial neural network based prediction of optimal pseudo-
damping and meta-damping in oscillatory fractional order dynamical systems, Advances in
Engineering, Science and Management (ICAESM), 2012 International Conference on, pp. 350-356,
Nagapattinam, Mar. 2012.
[29] M.D. Choudhury, S. Chandra, S. Nag, S. Das, and S. Tarafdar, Forced spreading and rheology
of starch gel: Viscoelastic modeling with fractional calculus, Colloids and Surfaces A:
Physicochemical and Engineering Aspects, vol. 407, pp. 64-70, Aug. 2012.
[30] A. Tepljakov, E. Petlenkov, and J. Belikov, FOMCON: fractional-order modeling and control
toolbox for MATLAB, Mixed Design of Integrated Circuits and Systems (MIXDES), 2011
Proceedings of the 18th International Conference, pp. 684-689, Gliwice, Jun. 2011.
[31] S. Das, I. Pan, S. Saha, A. Kumar, S. Das, and A. Gupta, Revisting Oustaloups recursive filter
for analog realization of fractional order differintegrators, Energy, Automation, and Signal (ICEAS),
2011 International Conference on, Bhubaneswar, Dec. 2011.
[32] D. Valrio, and J. S da Costa, Ninteger: a non-integer control toolbox for MatLab,
Proceedings of the First IFAC Workshop on Fractional Differentiation and Applications, Bordeaux,
France, 2004.
[33] M.S. Tavazoei and M. Haeri, Chaotic attractors in incommensurate fractional order systems,
Physica D: Nonlinear Phenomena, vol. 237, no. 20, pp. 2628-2637, Oct. 2008.
[34] X. Wu, H. Lu, and S. Chen, Synchronization of a new fractional-order hyperchaotic system,
Physics Letters A, vol. 373, no. 27-28, pp. 2329-2337, Jun. 2009.
[35] L. Molgedey, J. Schuchhardt, and H.G. Schuster, Suppressing chaos in neural networks by
noise, Physical Review Letters, vol. 69, no. 26, pp. 3717, Dec. 1992.
[36] C.-S. Poon, and M. Barahona, Titration of chaos with added noise, Proceedings of the National
Academy of Sciences, vol. 98, no. 13, pp. 7107-7112, Jun. 2001.
[37] J.B. Gao, C.C. Chen, S.K. Hwang, and J.M. Liu, Noise-induced chaos, International Journal of
Modern Physics B, vol. 13, no. 28,pp. 3283-3305, Nov. 1999.
[38] J.B. Gao, S.K. Hwang, and J.M. Liu, When can noise induce chaos? Physical Review Letters,
vol. 82, no. 6, pp. 1132, Feb. 1999.
[39] B. Dennis, R.A. Desharnais, J.M. Cushing, S.M. Henson, and R.F. Costantino, Can noise induce
chaos?, OIKOS, vol. 102, no. 2, pp. 329-339, Aug. 2003.
[40] S.P. Ellner, and P. Turchin, When can noise induce chaos and why does it matter: a critique,
OIKOS, vol. 111, no. 3, pp. 620-631, Dec. 2005.
[41] M.T. Rosenstein, J.J. Collins, and C.J. De Luca, A practical method for calculating largest
Lyapunov exponents from small data sets, Physica D: Nonlinear Phenomena, vol. 65, no. 1, pp. 117-
134, May 1993.
[42] Liu, Hai-Feng, Zheng-Hua Dai, Wei-Feng Li, Xin Gong, and Zun-Hong Yu, Noise robust
estimates of the largest Lyapunov exponent, Physics Letters A, vol. 341, no. 1, pp. 119-127, Jun.
2005.
[43] J.B. Gao, J. Hu, W.W. Tung, and Y.H. Cao, Distinguishing chaos from noise by scale-
dependent Lyapunov exponent, Physical Review E, vol. 74, no. 6, pp. 066204, Dec. 2006.

26
[44] M. Cencini, M. Falcioni, E. Olbrich, H. Kantz, and A. Vulpiani, Chaos or noise: Difficulties of a
distinction, Physical Review E, vol. 62, no. 1, pp. 427, Jul. 2000.
[45] W. Kinsner, Characterizing chaos through Lyapunov metrics, Systems, Man, and Cybernetics,
Part C: Applications and Reviews, IEEE Transactions on, vol. 36, no. 2, pp. 141-151, Mar. 2006.
[46] G.A. Gottwald and I. Melbourne, Testing for chaos in deterministic systems with noise,
Physica D: Nonlinear Phenomena, vol. 212, no. 12, pp. 100-110, Dec. 2005.
[47] U. Parlitz, Identification of true and spurious Lyapunov exponents from time series,
International Journal of Bifurcation and Chaos, vol. 2, no. 1, pp. 155-165, Mar. 1992.
[48] K. Geist, U. Parlitz, and W. Lauterborn, Comparison of different methods for computing
Lyapunov exponents, Progress of Theoretical Physics, vol. 83, no. 5, pp. 875-893, 1990.
[49] K. Ramasubramanian, and M. S. Sriram, A comparative study of computation of Lyapunov
spectra with different algorithms, Physica D: Nonlinear Phenomena, vol. 139, no. 1, pp. 72-86, May
2000.
[50] G.A. Gottwald, and I. Melbourne, A new test for chaos in deterministic systems, Proceedings
of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, vol. 460,
no. 2042, pp. 603-611, Feb. 2004.
[51] G.A. Gottwald, and I. Melbourne, On the implementation of the 0-1 test for chaos, SIAM
Journal on Applied Dynamical Systems, vol. 8, no. 1, pp. 129-145, 2009.
[52] I. Pan, S. Das, and A. Routh, Towards a global controller design for guaranteed synchronization
of switched chaotic systems, Applied Mathematical Modelling, vol. 39, no. 8, pp. 2311-2331, Apr.
2015.
[53] D. Liberzon, Switching in systems and control, Springer, Berlin, 2003.

27