You are on page 1of 6

On the search for short chaotic logistic map sequences with good

auto-correlation properties
Gerardo Laguna-Sanchez*, Daniela Aguirre-Guerrero, and Ismael Robles-Martinez
Universidad Autonoma Metropolitana, Mexico
November 09, 2021

Abstract variation on the sequence values. Since chaotic se-


quences look like random sequences, the latter may
This work presents short chaotic sequences with good be emulated by the former, with similar correlation
auto correlation properties. The studied sequences properties but with the advantage, for the chaotic
are synthesized by means of the logistic map function, sequences, of being generated in a deterministic and
and being chaotic but fully deterministic, become simple way.
a good alternative to well-known sequences, for ex-
Sequences with good correlation properties are
ample the Zadoff-Chu sequences, especially for small
very appreciated in many fields and are especially
size sequences commonly used for frame synchroniza-
important for synchronization purposes in the digital
tion purposes in digital communications systems. An
communications field [2]. Currently, mathematically
innovative auto-correlation goodness metric is pro-
defined sequences, such as the Zadoff-Chu sequences
posed, and it is employed as an objective function
are widely used for frame synchronization purposes,
to maximize the correlation goodness for small syn-
and their properties have been thoroughly analyzed
chronization sequences synthetized by means of the
[3]. Even, bio-inspired heuristics have been proposed
logistic map function.
to synthesize sequences with good auto correlation
properties [4].
1 Introduction In this work, the Chaotic Logistic Map (CLM) se-
quences are proposed as an alternative to generate
The logistic map is a mathematical function origi- sequences with good auto-correlation properties, by
nally derived for modeling the population size of a means of maximizing the proposed Peak Sidelobe at-
given ecosystem with constrictions that may present tenuation metric (PSA), used as an objective func-
a chaotic, but deterministic behavioral. Its general tion, over the search space defined by parameters
form is expressed by Equation 1. R and x(0). Although the application of the auto-

x(n + 1) = x(n) 1 − x(n) R (1) correlation properties of chaotic sequences for syn-
chronization and communications purposes is not a
The properties of the logistic map have been widely new idea [5], the main contribution of this work is to
studied [1] and it is well known that it generates a provide guidelines to determinate the optimum pa-
chaotic sequence when the R parameter is between rameters for the synthesis of chaotic sequences with
3.57 and 4. Furthermore, the logistic function is very good correlation properties.
sensitive to the initial conditions: a minimal differ- The rest of this work is presented as follows: In
ence in the initial value, x(0), produces a significant Section 2, the proposed Peak Sidelobe Attenuation
* g.laguna@correo.ler.uam.mx metric is introduced; in Section 3, the used experi-

1
mental methodology is explained; In Section 4, the This improved metric allows us to maximize the ra-
experimental results are presented; finally, in Section tio for maximum auto-correlation value with respect
5, the conclusions are drawn. to the Peak Sidelobe level, to maximize the synchro-
nization performance of a sequence, and this with
only a slight increase in computational cost over the
2 The proposed auto- PSL metric.
correlation goodness metric The use of specific sequences with good auto cor-
relation properties is usual for time synchronization
Some cross correlation and auto-correlation metrics in the digital communication systems. Specifically,
are typically computed for the sequences focusing on within the OFDM systems, the frame synchroniza-
their periodic or aperiodic performance.  For exam- tion is achieved taking advantage of Cyclic Prefixes
ple, let x = x(0), x(1), . . . , x(N − 1) denote a se- (CP), originally added to avoid Inter Symbol Interfer-
quence with N samples, where x(k) ∈ R, the peri- ence (ISI). In this kind of synchronization problems,
odic, c(k), and aperiodic, r(k), auto-correlations for the natural solution is the aperiodic auto-correlation
the sequence are defined as: computation to find the position of the cyclic prefix
and, thus, the beginning of the OFDM frame. Ad-
N −1
X ditionally, if the frame synchronization sequence is
c(k) = x(n)x(n + k)modN , 0 ≤ k ≤ (N − 1)
known, it also may be used for channel estimation
n=0
(2) purposes.

N
X −1 3 Experimental methodology
r(k) = x(n)x(n + k), 0 ≤ k ≤ (N − 1) (3)
n=0 In this work, the proposed Peak Sidelobe Attenua-
tion (PSA) is chosen as an objective function. Thus,
To evaluate the auto-correlation performance of a
the PSA metric allows us to explore its search space
given sequence, it is common to apply the following
to maximize the auto-correlation goodness for the
three metrics [6]:
chaotic sequences generated by Equation 1. Specif-
1. The Peak Sidelobe Level ically, for chaotic sequences generated by means of
the logistic map function, the useful search space is
−1
P SL = max{r(k)}N
k=1 , (4) defined by the R parameter range within (3.57, 4.0]
and x(0) range within (0.0, 1.0]. To outline the gen-
2. the Integrated Sidelobe Level eral shape of the objective function, the experimental
N −1
methodology is supported by a brute force algorithm.
X Basically, the algorithm executes the following steps,
ISL = x(n)2 , (5)
according to a specific resolution for both R and x(0)
k=1
ranges:
3. and the Merit Factor
r(0)2 r(0)2
MF = PN −1 = . (6)
2 k=1 r(k)2 2 × ISL

In this work, expressed in Equation 7, the Peak


Sidelobe Attenuation is introduced.

r(0)
P SA = . (7)
P SL

2
Data: Resolution and range for R and x(0)
parameters
Result: maximum PSA value
while not processed all pair of {R, x(0)} 1

parameters do 10
0.8

PSA [dB]
Generate a logistic map sequence with the 5 0.6

0.4

required size; 0

4
3.95
3.9
0.2
xo

Compute the auto-correlation sequence; 3.85


3.8

R
3.75
3.7
0

Compute and store the PSA metric;


end
Figure 1: 3D view of PSA function surface for search
Plot all the computed PSA values for the
space with 50x50 points.
search space;
Print the maximum PSA value on the search SPA [dB]
1

space and their related parameters; 8


0.9

Algorithm 1: Algorithm to find the maximum 7 0.8

PSA value on the search space. 6 0.7

0.6
5

xo
0.5
4

0.4

3
0.3

4 Experimental results and dis- 2

1
0.2

0.1

cussion 4 3.95 3.9 3.85


R
3.8 3.75 3.7
0

The experimental results are computationally very Figure 2: 2D view of PSA function surface for search
easy to reproduce. Here, just two representative cases space with 50x50 points.
are presented to outline the overall shape for the
search space (as surface in a 3D representation) that
must be used to maximize any auto-correlation good- ements, departing from the coordinates (R, x(0)) cor-
ness metric for chaotic sequences generated by means responding to a PSA local maximum with a 50 × 50
the logistic map function: a) search space with 50×50 low resolution, such as depicted in Figure 1 and Fig-
points resolution, and b) search space with 400 × 400 ure 2, PSA values around 8.68 dB are produced; while
points resolution. in the case of a 400 × 400 moderate resolution, as de-
In Figure 1 (3D view) and Figure 2 (2D view), picted in Figure 3 and Figure 4, PSA local maximum
the PSA surface is plotted, for logistic map chaotic values around 10.0 dB are produced. In fact, it was
sequences on the search space with a resolution of observed that PSA reaches maximum values around
50 × 50 points. In Figure 3 (3D view) and Figure 4 11.0 dB on search spaces with 100, 000 × 100, 000
(2D view), the PSA surface is plotted, on the same points resolution. In this work, since the used ex-
search space, but with a resolution of 400×400 points. perimental algorithm is based on a brute-force tac-
With different resolutions, both figures outline the tic, these optimization searches have a relatively high
overall shape for the PSA surface. computing cost and results in long time executions.
In Figures 1 to 4, it must be remarked that the av- This issue may be improved by using a heuristic op-
erage surface shows a waved shape with the highest timization technique (for example with evolutionary
PSA local maximums in the band defined by R with algorithms).
values between 3.96 and 4.0. Also, it was experimen- In Table 1, PSA local maximums are reported for
tally observed that the maximum PSA value achieved chaotic logistic map sequences with 64 elements over
increases with higher resolution of the search space. a search space with resolution of 100, 000 × 100, 000
For example, synthesizing chaotic sequences of 64 el- points.

3
Table 1: PSA metrics and parameters for some of
the best short (64 elements) chaotic logistic map se-
quences, found from the local PSA maximums over
search spaces with resolution of 100, 000 × 100, 000
points.
Figure 3: 3D view of PSA function surface for search PSA metric [dB] R value x(0) value
space with 400x400 points. 11.183 3.99119 0.06952
11.097 3.99894 0.17857
11.046 3.99986 0.20092
10.968 3.99881 0.36061
10.886 3.99913 0.47951
10.898 3.99926 0.53799
11.538 3.99871 0.67709
11.005 3.99099 0.71311
11.097 3.99894 0.82143
11.113 3.99807 0.90204

Figure 4: 2D view of PSA function surface for search


space with 400x400 points.

In Table 2, as a benchmark, the PSA metrics for


some of the best 64 elements Zadoff-Chu sequences
are reported. It must be remembered that the Zadoff-
Chu sequences, for an even number of elements, is Table 2: PSA metrics and M parameter for some of
defined by Equation 8, where N is the even size and the best 64 elements Zadoff-Chu sequences.
M is an integer relatively prime to N [7]. PSA metric [dB] M value
 M πn2  7.2312 5
z(n) = exp i . (8) 6.8756 7
N 7.4162 9
It is important to note that, for any small size se- 7.9959 13
quence, it is difficult to achieve a high value for any 8.6597 21
auto-correlation goodness metric. That is the main 8.2284 35
reason to choose 64 element sequences to compare 9.4423 43
the auto-correlation performance of the chaotic lo- 8.3289 51
gistic map sequences with respect to the well-known 7.9377 55
Zadoff-Chu sequences. Specifically, the LTE stan- 8.7702 63
dard specifies the use of small size sequences, with
only 63 elements, to get fine timing by means a cross-
correlation with the Primary Synchronization Signal
(PSS) (i.e., see [8] and the references therein).

4
To facilitate measuring the auto-correlation goodness
Auto-correlation performance

11.5 for sequences, the Peak Sidelobe Attenuation (PSA)


11 metric has been introduced. After experimental ex-
10.5
ploration of the related search space, to maximize
the PSA metric employed as an optimization objec-
10
tive function, it was observed that the PSA average
PSA [dB]

9.5
surface clearly shows a recognizable shape that al-
9
lows the optimization algorithms to find local maxi-
8.5 mums with the best auto-correlation goodness. The
8 maximum PSA values achieved, by chaotic logistic
7.5 map sequences with 64 elements and on search spaces
7
with resolution of 0.00001 for parameters R and x(0),
showed a significant improvement (twice as much on
CLM Zadoff-Chu
average) with respect to the auto-correlation good-
ness for the best Zadoff-Chu sequences with the same
Figure 5: Auto-correlation performance for both, length, which represents an evident advantage to be
CLM and Zadoff-Chu, 64 elements length sequences. successfully applied for both time and frame synchro-
nization processes.

After comparing the auto-correlation PSA good-


ness metrics, reported in Tables 1 and 2, the supe- References
riority of the chaotic logistic map (CLM) sequences
[1] R. M. May, “Simple mathematical models with
is clear, with respect to the Zadoff-Chu sequences.
very complicated dynamics,” Nature, vol. 261,
In Figure 5, the overall performance for PSA auto-
pp. 459–467, Jun. (1976).
correlation metric, for both CLM and Zadoff-Chu se-
quences, is depicted. It must be remarked than an av- [2] S. E. Kocabas and A. Atalar, “Binary sequences
erage 3dB PSA gain, of CLM sequences over the refer- with low aperiodic auto-correlation for synchro-
ence Zadoff-Chu ones, is achieved. These results show nization purposes,” Communication Letters, vol.
that, for the same length of 64 elements, the proposed 7, no. 1, pp. 36–38, Jan. (2003).
CLM sequences produce auto-correlations goodness
twice as high as the Zadoff-Chu ones, which is very [3] J. W. Kang, Y. Whang, H. B. Ko and K. S. Kim,
convenient to improve any time or frame synchroniza- “Generalized cross-correlation properties of Chu
tion process. Finally, considering that both sequences sequences,” IEEE Transactions on Information
are fully deterministic and synthesized by mathemat- Theory, vol. 58, no. 1, pp. 438–444, Jan. (2012).
ical expressions, the chaotic logistic map sequences
[4] B. Militzer, M. Zamparelli, D. Beule, “Evo-
become an interesting alternative to produce short
lutionary search for low autocorrelated binary
sequences with good auto-correlation properties.
sequences,” IEEE Transaction on Evolutionary
Computing, vol. 2, no. 1, pp. 34–39, Apr. (1998).
5 Conclusions [5] S. Rao and S. P. Howard, “Correlation perfor-
mance of chaotic signals in spread spectrum sys-
In this work, the logistic map function has been pro- tems,” in IEEE Digital Signal Processing Work-
posed as an alternative to synthesize short chaotic se- shop Proceedings, pp. 506–509 (1996).
quences, with good auto-correlation properties, that
show an auto-correlation goodness gain of 3dB over [6] M. Soltanalian and P. Stoica, “Computational
the best Zadoff-Chu sequences with the same length. Design of Sequences with good correlation prop-

5
erties,” IEEE Transactions on Signal Process-
ing, vol. 60, no. 5, pp. 2180–2193, May. (2012).
[7] D. C. Chu, “Polyphase codes with good periodic
correlation properties,” IEEE Transactions on
Information Theory, vol. 18, no. 4, pp. 531–532,
Jul. (1972).
[8] A. Donarski, T. Lamahewa and J. Sorensen,
“Downlink LTE synchronization: A software de-
fined radio approach,” in 8th International Con-
ference on Signal Processing and Communica-
tion Systems (ICSPCS), pp. 1–9, (2014).

You might also like