Analysis of Chaotic Behavior in

Non-linear Dynamical Systems Micha■

Piórek
Visit to download the full and correct content document:
https://textbookfull.com/product/analysis-of-chaotic-behavior-in-non-linear-dynamical-
systems-michal-piorek/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Non-Linear Differential Equations and Dynamical Systems

1st Edition Luis Manuel Braga Da Costa Campos

https://textbookfull.com/product/non-linear-differential-
equations-and-dynamical-systems-1st-edition-luis-manuel-braga-da-
costa-campos/

Intelligent Algorithms for Analysis and Control of

Dynamical Systems Rajesh Kumar

https://textbookfull.com/product/intelligent-algorithms-for-
analysis-and-control-of-dynamical-systems-rajesh-kumar/

Dynamical Systems in Population Biology Xiao-Qiang Zhao

https://textbookfull.com/product/dynamical-systems-in-population-
biology-xiao-qiang-zhao/

Dynamical systems in population biology Second Edition

Zhao

https://textbookfull.com/product/dynamical-systems-in-population-
biology-second-edition-zhao/
Non-cooperative Stochastic Differential Game Theory of
Generalized Markov Jump Linear Systems 1st Edition
Cheng-Ke Zhang

https://textbookfull.com/product/non-cooperative-stochastic-
differential-game-theory-of-generalized-markov-jump-linear-
systems-1st-edition-cheng-ke-zhang/

Spaces of dynamical systems 2nd Edition Pilyugin S.Yu

https://textbookfull.com/product/spaces-of-dynamical-systems-2nd-
edition-pilyugin-s-yu/

Spaces of Dynamical Systems 2nd Edition Sergei Yu.

Pilyugin

https://textbookfull.com/product/spaces-of-dynamical-systems-2nd-
edition-sergei-yu-pilyugin/

Analysis and Data Based Reconstruction of Complex

Nonlinear Dynamical Systems Using the Methods of
Stochastic Processes M. Reza Rahimi Tabar

https://textbookfull.com/product/analysis-and-data-based-
reconstruction-of-complex-nonlinear-dynamical-systems-using-the-
methods-of-stochastic-processes-m-reza-rahimi-tabar/

Linear Systems Signal Processing and Hypercomplex

Analysis Chapman University November 2017 Daniel Alpay

https://textbookfull.com/product/linear-systems-signal-
processing-and-hypercomplex-analysis-chapman-university-
november-2017-daniel-alpay/
Studies in Systems, Decision and Control 160

Michał Piórek

Analysis of Chaotic
Behavior in Non-
linear Dynamical
Systems
Models and Algorithms for Quaternions
Studies in Systems, Decision and Control

Volume 160

Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: kacprzyk@ibspan.waw.pl
The series “Studies in Systems, Decision and Control” (SSDC) covers both new
developments and advances, as well as the state of the art, in the various areas of
broadly perceived systems, decision making and control-quickly, up to date and
with a high quality. The intent is to cover the theory, applications, and perspectives
on the state of the art and future developments relevant to systems, decision
making, control, complex processes and related areas, as embedded in the fields of
engineering, computer science, physics, economics, social and life sciences, as well
as the paradigms and methodologies behind them. The series contains monographs,
textbooks, lecture notes and edited volumes in systems, decision making and
control spanning the areas of Cyber-Physical Systems, Autonomous Systems,
Sensor Networks, Control Systems, Energy Systems, Automotive Systems,
Biological Systems, Vehicular Networking and Connected Vehicles, Aerospace
Systems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power
Systems, Robotics, Social Systems, Economic Systems and other. Of particular
value to both the contributors and the readership are the short publication timeframe
and the world-wide distribution and exposure which enable both a wide and rapid
dissemination of research output.

More information about this series at http://www.springer.com/series/13304

Michał Piórek

Analysis of Chaotic Behavior

in Non-linear Dynamical
Systems
Models and Algorithms for Quaternions

123
Michał Piórek
Department of Computer Engineering
Wrocław University of Science
and Technology
Wrocław, Poland

ISSN 2198-4182 ISSN 2198-4190 (electronic)

Studies in Systems, Decision and Control
ISBN 978-3-319-94886-7 ISBN 978-3-319-94887-4 (eBook)
https://doi.org/10.1007/978-3-319-94887-4
Library of Congress Control Number: 2018946568

© Springer International Publishing AG, part of Springer Nature 2019

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by the registered company Springer International Publishing AG
part of Springer Nature
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To my Wife and my Family
Acknowledgements

Many people have contributed to this work. At first, I would like to express my
deepest gratitude to Prof. Ewa Skubalska-Rafajłowicz and Dr. Bartosz Jabłoski for
the countless support during the study on chaotic systems and quaternions. Their
guidance helped me in all the time of research and writing of this book.
I would like to express my appreciation to Prof. Józef Korbicz and Prof. Andrzej
Polański for the comments, which helped me to increase the value of this work.
I would also like to thank Prof. Ewaryst Rafajłowicz, Prof. Tyll Krueger, and Dr.
Łukasz Korus for very important advice and help during the research.
My sincere thanks also go to Prof. Konrad Wojciechowski, Dr. Henryk Josiński,
Dr. Adam Świtoński, and colleagues from Silesian University of Technology, who
provided me an opportunity to work with their team, and who gave access to the
great gait recordings data set. Without their precious support, it would not be
possible to conduct this research. Human motion data used in the experiments
presented in this book was obtained from the Centre for Research and Development
of Polish-Japanease Academy of Information Technology (www.bytom.pja.edu.pl).
A special thanks to my family, my wife for her patience and indulgence, my
parents, siblings, and friends for supporting me spiritually. I hope we will have
more time for each other now.

vii
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 1
1.1 Problem Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 3
1.2 Proposed Methods and Algorithms, Carried Out Numerical
Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Novelties Presented in This Work . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Book Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Processes Described by Quaternion Models . . . . . . . . . . . . . . . . . . 7
2.1 Definition of the Processes Described by Quaternion Models . . . 7
2.2 Elements of Quaternion’s Algebra . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Quaternions Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Quaternions Averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Quaternions Random Generation . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6 Advantages and Disadvantages of Using Quaternions
Parametrization of Rotation . . . . . . . . . . . . . . . . . . . . . . . . . ... 16
3 Deterministic Chaos Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1 Dynamical Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Chaos Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.1 Positive Entropies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.2 Strong Sensitivity to Initial Conditions . . . . . . . . . . . . . . 26
3.2.3 Strange Attractor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.4 Non-integer Fractal Dimension of the Attractor . . . . . . . . 29
3.3 Analysis of Chaos Basing on a Time Series . . . . . . . . . . . . . . . . 30
3.4 Time Delay Embedding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4 Analysis of Chaos from Time Series - Existing Methods Survey . . . 35
4.1 Time Delay . . . . . . . . . . . . . . . . . . . . . . . ................ 35
4.2 Embedding Dimension . . . . . . . . . . . . . . . ................ 37
4.3 Reconstruction of the Phase Space . . . . . . . ................ 38

ix
x Contents

4.4 The Largest Lyapunov’s Exponent . . . . . . . . . . . . . . . . . . . . . . . 40

4.5 Entropies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.6 Fractal Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5 Analysis of Chaos from Quaternion Time Series - Proposed
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.1 Quaternion’s Angle Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.2 Time Delay Embedding for Quaternion Time Series . . . . . . . . . . 47
5.3 Mutual Information for Quaternions . . . . . . . . . . . . . . . . . . . . . . 48
5.4 Quaternions Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.5 Quaternions Clustering Validity Measures . . . . . . . . . . . . . . . . . 53
5.5.1 Quaternion Davies–Bouldin Index (QDB) . . . . . . . . . . . . 54
5.5.2 Quaternion Dunn’s Index (QDI) . . . . . . . . . . . . . . . . . . . 54
5.5.3 Quaternion Calinski–Harabasz Index (QCH) . . . . . . . . . . 56
5.6 False Nearest Neighbours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.7 The Largest Lyapunov’s Exponent . . . . . . . . . . . . . . . . . . . . . . . 60
5.8 Correlation Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6 Numerical Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.1 Experiments Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.2 Investigated Time Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.3 Gait Quaternion Time Series . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.4 Random Quaternion Time Series . . . . . . . . . . . . . . . . . . . . . . . . 70
6.5 Periodic Quaternion Time Series . . . . . . . . . . . . . . . . . . . . . . . . 70
6.6 The Aim of Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
7 Analysis of Chaos for Quaternion Time Series . . . . . . . . . . . . . . . . 73
7.1 Analysis of Chaos - Gait Time Series . . . . . . . . . . . . . . . . . . . . 73
7.1.1 Clusters Number Estimation . . . . . . . . . . . . . . . . . . . . . . 73
7.1.2 Time Delay Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.1.3 Embedding Dimension Estimation . . . . . . . . . . . . . . . . . 75
7.1.4 Phase Space Reconstruction . . . . . . . . . . . . . . . . . . . . . . 76
7.1.5 The Largest Lyapunov’s Exponent Estimation . . . . . . . . . 76
7.1.6 Correlation Dimension Estimation . . . . . . . . . . . . . . . . . . 77
7.2 Analysis of Chaos - Periodic Time Series . . . . . . . . . . . . . . . . . 78
7.2.1 Clusters Number Estimation . . . . . . . . . . . . . . . . . . . . . . 78
7.2.2 Time Delay Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.2.3 Embedding Dimension Estimation . . . . . . . . . . . . . . . . . 80
7.2.4 Phase Space Reconstruction . . . . . . . . . . . . . . . . . . . . . . 80
7.2.5 The Largest Lyapunov Exponent Estimation . . . . . . . . . . 81
7.2.6 Correlation Dimension Estimation . . . . . . . . . . . . . . . . . . 81
7.3 Analysis of Chaos - Random Time Series . . . . . . . . . . . . . . . . . 82
7.3.1 Clusters Number Estimation . . . . . . . . . . . . . . . . . . . . . . 82
7.3.2 Time Delay Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Contents xi

7.3.3 Embedding Dimension Estimation . . . . . . . . . . . . . . . . . 84

7.3.4 Phase Space Reconstruction . . . . . . . . . . . . . . . . . . . . . . 84
7.3.5 The Largest Lyapunov Exponent Estimation . . . . . . . . . . 85
7.3.6 Correlation Dimension Estimation . . . . . . . . . . . . . . . . . . 86
7.4 Experiments Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
8 Comparison Against Existing Approaches . . . . . . . . . . . . . . . . . . . 89
8.1 Compared Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.2 Medical Angles Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
8.2.1 Medical Angles Analysis Procedure . . . . . . . . . . . . . . . . 91
8.2.2 Model Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.2.3 LLE Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
8.3 Quaternion Angles Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.3.1 Quaternion Angles Embedding Procedure . . . . . . . . . . . . 96
8.3.2 Model Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.3.3 LLE Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
8.4 Quaternions Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
8.4.1 Quaternions Analysis Procedure . . . . . . . . . . . . . . . . . . . 100
8.4.2 Clusters Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
8.4.3 Model Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.4.4 LLE Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.5 Experiments Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
9 Quaternions Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
9.1 Optimal Clusters Number Selection - Data Sets Generated
Around Known Number of Clusters . . . . . . . . . . . . . . . . . . . . . . 111
9.2 Optimal Clusters Number Selection - Data with Varied
Overlapping Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
9.3 Optimal Clusters Number Selection - Data with Different
Variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
9.4 Experiments Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Chapter 1
Introduction

Deterministic chaos is a irregular behaviour, which from some perspective might

appear to be random but in reality it is driven by a deterministic rules. From math-
ematical perspective it is a set of properties characterizing the dynamical systems -
mathematical models describing the evolution of some phenomena in time. Once a
particular dynamical system preserves that set of properties it can be called chaotic.
Chaosity of the dynamical system is mostly caused by it’s complexity. Hence,
very often it is observed in the case of nonlinear dynamical systems, which have
more complex nature than linear. Majority of natural processes have got nonlinear
nature and since that, they are modeled by nonlinear dynamical systems, which can
exhibit chaotic properties. This is the root cause of popularity of analysis of chaos
in the real systems.
The deterministic chaos phenomena has been observed so far in meteorology
[121, 138], economy [104], road traffic [98, 118], medicine [2, 141], biology [126],
astronomy [111] and sociology [77].
While analyzing a chaos properties of real systems the exact form of the dynam-
ical system (e.g. N -first order ordinary differential equations) is usually not known.
Hence, analysis is performed on the time series, which is a collection of measurements
of one of the outputs of the system in time. This implies a fact that, a confirmation
of the occurrence of deterministic chaos in the analyzed system requires a certain set
of methods designed for time series.
Usually it can be done by the approach, which includes several steps:
1. Model of the system selection
2. Reconstruction of the phase space using the chosen model
3. Estimation of quantities, which may confirm existence of deterministic chaos
e.g. sensitivity to initial conditions, fractal structure of the attractor or system’s
entropy

© Springer International Publishing AG, part of Springer Nature 2019 1

M. Piórek, Analysis of Chaotic Behavior in Non-linear Dynamical Systems,
Studies in Systems, Decision and Control 160,
https://doi.org/10.1007/978-3-319-94887-4_1
2 1 Introduction

This methodology is called a phase space approach of nonlinear analysis of time

series. It assumes that examined time series representing the knowledge about the
real system is an effect of underlying dynamical system evolution. Basing on the
above steps one may judge about the chaotic properties of the examined process.
The studies on deterministic chaos have begun since the work on weather predic-
tion published by Edward Lorenz in [91]. The fundamental work opening the field of
reconstruction of the phase space of the dynamical system from the measurements
of the system’s output was published by Floris Takens in [132]. It is called Takens
Theorem and it started a massive interest in the field of methods of analysis of chaos
in the real systems. The theorem was initially proofed for autonomous systems, then
it was extended to systems with forcing [129] and systems with stochastic forcing
[128].
Since the publication of Takens theorem a lot of related work have been published
regarding the analysis of chaos basing on scalar time series (where each sample of
the time series is a single real value). The wealth of the methods (elaborated in the
Chap. 4) for scalar time series is not transferred into the case when the subject time
series is a multivariate one, especially when the input time series is a quaternion time
series.
Recently, a lot of effort is being done on the field of quaternions analysis. Unit
quaternions - a subgroup of quaternions space are utilized for parametrization of
the rotation of a body in 3D space. Their advantage against of matrices of rotations
and Euler’s angles parametrization are a lack of singularities while describing the
rotation and computational complexity reduction while calculating rotations. This
convenience allows to use quaternions in the aerospace or underwater attitude control
[28, 45, 83] or in motion capture data processing [64, 147].
Apart from describing rotations of the body in the 3D space, quaternions can also
be used for representation of a color in the RGB image and applied in the image
processing field e.g. image classification [112], color image transformation [41], 3D
face landmark labeling [32] or image principal component analysis [85].
Very promising quaternion usage is also related to optic flows [9, 11, 29] and
diagnostics of medical surfaces [30, 145].
Therefore the application of quaternions is so wide, the natural question arrived
while working on quaternions analysis: is it possible to observe the deterministic
chaos property in the processes described by quaternions?
To answer this question a new set of methods of analysis of chaos in the processes
described by quaternion models needs to be incorporated. As per the established
literature research and Author’s knowledge, recently, there is no active direction of
research on this subject in the related work. This implies the field itself seems to be
new and a set of new methods needs to be proposed.
All of the methods and results presented in this book are an outcome of experi-
ments and studies on rotations sequences especially human gait recordings.
1.1 Problem Characteristics 3

1.1 Problem Characteristics

The main problem elaborated in this book can be defined as: Propose methods and
algorithms of nonlinear analysis of processes described by quaternion models allow-
ing for the detection of deterministic chaos presence.
Having that methodology in place one may answer the question whether basing
on the quaternion time series created from observations of the dynamical system
output is it possible to detect the deterministic chaos in the system?
For the performed research and the proposed methods the below assumptions
have been made:
1. As processes defined by quaternion models we treat the processes, which are ruled
by dynamical systems evolution and their output can be recorded in quaternion
parametrization form, especially quaternion time series.
2. In the scope of the research are stationary, autonomous dynamical systems, with-
out forcing.
3. The model of the system is created by time delay embedding. We consider in this
work the rotations space (a subset of R n space), which satisfies Takens embedding
theorem.

1.2 Proposed Methods and Algorithms, Carried Out

Numerical Experiments

The book depicts a 5 years of studies on the deterministic chaos in the real processes,
which can be described by quaternions. The results of presented research can be
divided into two kinds: methodology and experiments.
From methodological point one may list the following algorithms have been pro-
posed:
• Mutual information method for quaternion time series - method for estimation of
time delay.
• False nearest neighbors method for quaternion time series - method for estimation
embedding dimension.
• The largest Lyapunov exponent estimation algorithm for quaternion domain.
• Correlation dimension estimation algorithm for the attractor in quaternion domain.
• kMeans clustering algorithm for quaternion data sets.
• Quaternions clustering validity indexes.
From the experimental point, validity and properties of presented methods have
been addressed in the experiments:
• Experiment of usage of proposed methods to differentiate random, chaotic and
periodic quaternion time series.
4 1 Introduction

• Experiment of comparison proposed chaos analysis approach against univariate

approaches for gait sequences analysis.
• Experiments of testing quaternions clustering framework against different type of
input data sets.

1.3 Novelties Presented in This Work

The research presented in this book can be considered as a new one in the field
because of several reasons:
• As per the literature research, in the related work there is no existing direction of
analysis of chaos in the processes described by quaternions.
• Provided methods allow not only for a detection of deterministic chaos in quater-
nion time series but also for the classification of the nature of dynamical system,
which drives the quaternion time series. Using the proposed methodology three
types of the quaternion time series can be identified: random, periodic and chaotic.
• Proposed methods have been used to establish an analysis of chaos in real process -
gait sequences. The results of analysis for wide input data set have been compared
against the ‘classic’ methods designed for scalar time series.
• The additional value of the book is brought by the proposed framework of quater-
nions clustering. The framework includes the proposition of quaternions clustering
algorithm along with the measures of the quaternions clustering validity.
• Quaternions clustering framework has been tested on generated test data sets. The
quality of quaternions clustering has been tested against three parameters of the
input data set: real clusters number, overlapping degree and variance of the data
set.

1.4 Motivations

The entry point of analysis of chaos in the processes described by quaternion models
was an analysis of human gait recordings. Initially work was directed on nonlinear
analysis of gait time series in the context of human gait stability. The Author had
a great opportunity to work on this matter with the research team from Silesian
University of Technology, the owner of the massive gait recordings set recorded
using motion capture technique by system Vicon.
First experiments were performed using classic methods of embedding for a uni-
variate time series. Having the full information of recorded rotation in time (e.g
Euler angles or unit quaternions), only one-dimensional subset of it was taken into
consideration e.g. time series consisted of rotations in one plane or quaternions angle
time series. At some point of time there has arisen a question whether the analysis of
the full rotation description will bring an advance? Hence the work on embedding of
1.4 Motivations 5

unit quaternions has started since the unit quaternion representation is an equivalent
of Euler angles one.
The problem tends to be more general motivation to answer the questions whether
embedding of multivariate time series can bring an advance against embedding on
univariate time series and whether deterministic chaos can occur in the processes
described by quaternions?

1.5 Book Structure

Research presented in the book is organized as following.

Second chapter defines processes described by quaternion models and gathers
used quaternions algebra aspects. It also includes the random quaternion generation
approach and quaternions visualization approach descriptions.
Chapter three recalls the theory of dynamical systems. It also elaborates a deter-
ministic chaos properties, time delay embedding theory and concept of analysis of
chaos basing on a time series.
A survey of existing methods of analysis of chaos in a univariate time series is
presented in the chapter fourth. Apart from the most popular methods, a lot of other
available in the related work methods have been mentioned.
In the most important chapter number five all proposed methods for analysis
of chaos in the processes described by quaternion models have been presented. It
includes also quaternions clustering framework description.
First of experiments, analysis of chaos for three types of quaternion time series:
chaotic, random and periodic has been described in chapter sixth.
Experiment of comparison of analysis of chaos using proposed methods against
other available approaches for large data set from real system (gait sequences) is
presented in chapter seventh.
In chapter eight the experiments performed on the proposed clustering framework
have been depicted.
Conclusions and remarks have been gathered in the summary in the chapter ninth.
Chapter 2
Processes Described by Quaternion
Models

2.1 Definition of the Processes Described by Quaternion

Models

At the very beginning, to make the further reading, analysis and experiments clearer
we will focus on the definition of the processes described by quaternion models as
this is in the title of the book. The author would like to emphasize that any new
mathematical concept of quaternion processes is introduced here.
We assume the processes described by quaternion models are processes, which are
ruled by dynamical systems evolution and their output can be measured in quaternion
parametrization form, especially quaternion time series. Since that, it is crucial to
recall the theory of dynamical systems. This will be a subject of considerations of
next chapter.
The dynamical system concept includes three main components: phase space,
time and the time evolution law. A phase space X represents the possible states of
the system. Time can be a continuous variable or a discrete integer-valued variable.
The time evolution law, is a rule that allows us to determine the state of the system
at each moment of time t from its states at all previous times. An example of a
dynamical system is N first-order, autonomous, ordinary differential equations:

x (1) /dt = F1 (x (1) , x (2) , . . . , x (N ) ),

x (2) /dt = F2 (x (1) , x (2) , . . . , x (N ) ),
.. (2.1)
.
x (N ) /dt = FN (x (1) , x (2) , . . . , x (N ) )

We will perform all observations and analysis of such kind of processes on a time
series recorded as samples measured in the output of the underlying system.

s(n) = F(x(t0 + nΔt)), (2.2)

© Springer International Publishing AG, part of Springer Nature 2019 7

M. Piórek, Analysis of Chaotic Behavior in Non-linear Dynamical Systems,
Studies in Systems, Decision and Control 160,
https://doi.org/10.1007/978-3-319-94887-4_2
8 2 Processes Described by Quaternion Models

where s(n) is the value of nth sample of measurement of physical process x in time
t0 + nΔt and t0 is initial time.
In the presented research we will focus only on the processes, which are results of
those systems, for which the output can be recorded in quaternions parametrization
form. This is a novel approach and ultimately defines the scope of the processes
described by quaternion models. Hence, the main subject of further analysis will be
a unit quaternions time series representing the knowledge about the processes:

Q(n) = (q1 , q2 , . . . , q N )
(2.3)
= (w1 + i x1 + j y1 + kz 1 , . . . , w N + i x N + j y N + kz N )

where w is quaternion’s real unit and i, j, k are quaternion’s imaginary units.

This approach is designed mainly to rotational data and opens a field for a nonlinear
analysis of chaotic properties for a rotational data (especially in unit quaternion
parametrization form).
All developed and presented in this work methods are designed for a quaternion
domain, in which investigation of deterministic chaos properties will take place. For
the purpose of further considerations of algorithms working in a quaternions domain
we will rewind the fundamentals of quaternions algebra.

2.2 Elements of Quaternion’s Algebra

Quaternions are computationally efficient parametrization of rotational data. They

are hyper complex numbers (an extension of complex numbers) defined as following:

q = [w, ρ] = [w, (x, y, z)] = w + i x + j y + kz (2.4)

where: w represents a real part and ρ = (x, y, z) is called a vector part (i, j and k
are equivalents of imaginary unit). The whole quaternion set is usually denoted as
H , which is due to it’s main founder Sir William Hamilton.
The details of quaternions algebra widely used in the parametrization of rotations
are well described in the related work (e.g. [5, 31, 56, 65, 84, 139] or [62]).
In the scope of our interests are unit quaternions which can parametrize the space
of rotations:
q = 1 (2.5)

where quaternion norm is defined by:


q = w2 + x 2 + y 2 + z 2 (2.6)
2.2 Elements of Quaternion’s Algebra 9

Unit quaternions set is a subset of quaternion set H usually denoted as H1 . Unit

quaternions can be treated as a unit vectors in R 4 space and hence that the set of
whole unit quaternions creates a quaternion hypersphere.
Each unit quaternion q ∈ H1 can be represented in polar form:
 →

q = cosϕ, n sinϕ , (2.7)

→
where n ∈ R 3 is called an axis of the rotation and ϕ is called an angle of the rotation.
Quaternion conjugation is given as:

q ∗ = [w, (−x, −y, −z)] = w − i x − j y − kz (2.8)

Quaternion reciprocal is described by:

q∗
q −1 = (2.9)
q2

where reciprocal is defined for each q ∈ H \[0, (0, 0, 0)]. Quaternion multiplication
is defined as:

q1 xq2 = (w1 + x1 i + y1 j + z 1 k)(w2 + x2 i + y2 j + z 2 k)

= w1 w2 − x1 x2 − y1 y2 − z 1 z 2
+ (w1 x2 + x1 w2 + y1 z 2 − z 1 y2 )i (2.10)
+ (w1 y2 − x1 z 2 + y1 w2 + z 1 x2 ) j
+ (w1 z 2 + x1 y2 − y1 x2 + z 1 w2 )k

Operation of the quaternion’s logarithm, which is a setting the tangent space in the
point [1, (0, 0, 0)], is denoted by:
 ϕ 
logq = 0, n (2.11)
2
where n < π .
Let us suppose that r = (x, y, z) is a point in 3D space and qr represents the same
vector in quaternions form:
→
qr = [0, r ] = 0 + i x + j y + kz (2.12)

The rotated vector qr can be calculated by quaternion multiplication of:

qr = qqr q −1 , (2.13)

10 2 Processes Described by Quaternion Models

Fig. 2.1 Quaternion

parametrization of the
rotation - concept [122]

where q defines rotation angle and axis, as it was stated in Eq. 2.7. The whole
procedure is depicted by the concept diagram in the Fig. 2.1.
All time series considered in this book are unit quaternions time series and con-
cern of parametrization of rotations. All 3D rotations form S O(3), which is special
orthogonal group. Quaternions can be considered as one of parametrization of this
group.
The algorithms proposed in this work in most cases employ the distance function
between the two rotations parametrized by unit quaternion. Hence, it is important
to have well defined distance function (metric) between two elements of a data set.
Any distance function d used for handling rotation data, need to satisfy both usual
axioms for metrics and S O(3) properties [61]:
• d(R1 , R2 ) = 0 ↔ R1 = R2
• d(R1 , R2 ) = d(R2 , R1 )
• d(R1 , R3 ) ≤ d(R1 , R2 ) + d(R2 , R3 )
• function d respects the topology of S O(3)
• function d is left/right invariant or bi-invariant
where function d(R1 , R2 ) we call bi-invariant if d(R1 ∗ R2 , R1 ∗ R3 ) = d(R2 , R3 )
and d(R2 ∗ R1 , R3 ∗ R1 ) = d(R2 , R3 ).
A good survey of available quaternions distance functions for S O(3) can be found
in [61]. All of them are gathered in the Table 2.1.
The distance measure commonly used for the purposes of this research is d6
(functionally equivalent to d3 ). It has all required properties (metric axioms, bi-
variance, respecting S O(3) topology). It was used in some of the related papers [63,
64] and provided very good results in practical applications. The physical interpreta-
tion of this measure is a length of geodesic curve on the quaternion unit sphere. It can
be also considered as the amount of energy or rotation needed to rotate quaternion
q1 to the rotation defined by quaternion q2 .
2.3 Quaternions Visualization 11

Table 2.1 Quaternions distance functions

Name
Euclidean Distance between angles
Difference of Quaternions
Inner Product
Inner Product without inverse cosine
Deviation from the Identity Matrix
Geodesic on the Unit Sphere
Distance function
d1 ((α1 , β1 , γ1 ), (α2 , β2 , γ2 )) =
d(α1 , α2 )2 + d(β1 , β2 )2 + d(γ1 , γ2 )2
d2 (q1 , q2 ) = min{||q1 − q2 ||, ||q1 + q2 ||}
d3 (q1 , q2 ) = arccos(|q1 · q2 |)
d4 (q1 , q2 ) = 1 − |q1 · q2 |
d5 (R1 , R2 ) = ||I − R1 R2T || F
d6 (R1 , R2 ) = || log(R1 R2T )||

2.3 Quaternions Visualization

While designing the algorithms for analysis of chaos in the quaternions domain, it
is very beneficial to visualize their output to see the physical results of work. Proper
quaternions time series visualization can increase the understanding and intuition
about the investigated matter.
There are a few quaternion’s visualization techniques, they are presented in
[56, 57] or [68]. In the case of this book, the subject of analysis are rotational data
so the visualization approach should highlight the rotation direction instead of just
representing the unit quaternion. In the light of this, the author sees the visualiza-
tion proposed by Dr. Bartosz Jabłoński in his works [65] and [67] as a best choice.
The Author would like to thank Dr. Bartosz Jabłoński for the access to his software
implementing mentioned visualization approach and possibility to plot quaternions
data figures presented in this book.
The selected visualization is common for all types of parametrization for rotational
space, it can present the rotations written in unit quaternions form as well as Euler
angles and rotation matrix. The data originally recorded as Euler angles are presented
using selected method in the Fig. 2.2.

Fig. 2.2 The sphere of

rotation concept to visualize
rotational datasets [63]
12 2 Processes Described by Quaternion Models

The selected concept utilizes directed arrows representing data items. For each
data point the placement and orientation of the corresponding arrow describes the
rotation. Each data point is visualized as a series of three consecutive rotations around
axes (x, y, z).
In the Fig. 2.2 the gray arrow represents reference (base) orientation. The red
arrow depicts the rotation of base orientation around X ax. Similarly green and blue
arrows depict the analogical rotations of basis point around Y and Z axes. A bit more
complex rotation is depicted by the yellow element.
In the case of presence a massive data set it might be beneficial to replace the
arrows with dots for the better readability.

2.4 Quaternions Averaging

Very crucial aspect of the methodology proposed in this research is also a selection
of quaternion averaging method employed in algorithms. Apart from selection, very
important is to be aware of advantages and disadvantages of used method. A good
survey of the available methods is presented in [93]. The motivation of Markley’s
paper was to propose the optimal weighted unit quaternions average. In case of the
algorithms presented in this work it is sufficient to have non-weighted average but
having the weighted one algorithm is more than comfortable since one can always
set all the weights to one.
The simplest procedure of estimation weighted quaternions average can be defined
as:  n −1 n
 
q simple = wi wi qi (2.14)
i=1 i=1

This approach has two consequences. The first one is the estimated average is not a
unit quaternion, however this can be easily fixed by dividing quaternion by its norm.
The second is more important: q and −q represent the same rotation and hence that
there is 2:1 mapping over quaternion group. In result changing the sign of any qi
shouldn’t change the average value but Eq. 2.14 does not hold this property.
As it was found out in the paper [100], the problem of averaging of quaternions
can be solved with help of attitude matrices description. The attitude matrix is related
to the quaternion by the formula:

A(q) = w2 − ρ2 I3X 3 + 2ρρ T − 2w[ρ X ] (2.15)

where ρ and w are a vector and scalar parts of a quaternion (Eq. 2.4), I3 is a 3X 3
identity matrix and [ρ X ] is a cross product matrix defined by
⎡ ⎤
0 −z y
[ρ X ] = ⎣ z 0 −x ⎦ (2.16)
−y x 0
2.4 Quaternions Averaging 13

Following this observation Markley’s algorithms minimizes a weighted sum of

squared Forbenius norms of attitude matrix differences


n
q Mar kley = arg min wi A(q) − A(qi )2F (2.17)
q∈S3 i=1

where S3 denotes unit 3 sphere.

To see the differences between q Mar kley and q simple approaches we will com-
pute the averages using both algorithms for a small-length unit quaternion data set
Q example . The analyzed data set consists of the unit quaternions created from the
following rotations:

Q example = {(0, 0, 30); (0, 0, 60); (−30, 0, 60); (30, 0, 60); (0, 0, 90);
(2.18)
(0, 0, −30); (0, 0, −60); (−30, 0, −60); (30, 0, −60); (0, 0, −90)}

For this short quaternion data set two averages have been calculated and com-
pared in the Table 2.2. First average was computed using Q simple algorithm and then
normalized. The second average was computed using Q Mar kley . The example time
series along with computed average rotations are also presented on the unit quater-
nions sphere in the Fig. 2.3. All rotations form the sample set are marked as violet
when the both averages are marked as pink and light blue.
From this simple comparison one may see the quaternion averages calculated
using two algorithms represent the same rotation: both occurred on the north pole
of unit quaternion sphere. The averages are equal to the value but with different
sign. Moreover, the Markley’s algorithm seems to provide the average with some

Table 2.2 Quaternions averages algorithms comparison

Q simple Q simple_nor mali zed Q Mar kley
0.8424 + 0i + 0 j + 0k 1 + 0i + 0 j + 0k −1 − (1.11 ∗ 10−16 )i + 0 j + (1.11 ∗ 10−16 )k

Fig. 2.3 Visualization of the

small data set’s averages
14 2 Processes Described by Quaternion Models

numerical error, which is not relevant in the presented test case but can result with
higher numerical discrepancies in further quaternions processing (e.g. when the
quaternions time series is longer or the average is being computed multiple times
in one algorithm). This numerical error can be resolved during normalization of
algebraic mean. For the purposes of the algorithms presented in the book the simple
algorithm has been used. However, the author sees the testing of the application of
different quaternions averages against its performance can be a interesting subject
for the further research.

2.5 Quaternions Random Generation

A selection of quaternions random generation algorithm is also a prerequisite for

the implementation and testing of some algorithms described in this work. The main
task for such kind of algorithm is to generate normally distributed quaternions data
sets, which are additionally concentrated around a few centers(we can consider this
as a clusters centers). The input parameters given to the algorithm should be: clusters
centers, length of the output data set and distribution variance.
A survey of quaternions random variables definitions can be found in the related
work [90]. For the purposes of the current work the author selected the approach
proposed by Johnson in [65], which is called wrapping approach. We assume, the
multivariate non-degenerate normal distribution in k-dimensional case is defined as
 
1 1
f x (x, μ) =  exp − (x − μ)T K −1 (x − μ) , (2.19)
(2π )k |K | 2

where K is positive definite symmetric covariance matrix.

The foundation of the wrapping approach is that the random quaternion element
is calculated as a regular zero mean multivariate normal distribution in tangent space
T S 3 rotated at selected qm quaternion representing distribution mean. The visualiza-
tion of probability distribution function for quaternions generated using this approach
is presented in the Fig. 2.4.
The above model can be simplified. Using quaternion properties one may obtain
the following formula for probability density function:
 
1
p(q) = c exp − log (q M ∗ q)T K −1 log (q M ∗ q) , (2.20)
2

where c is a normalization parameter and K is covariance matrix.

In the experiments presented in this book, we will make use of simplified random
model with fixed variance at σq value for all dimensions.
The example usage of the algorithm is presented in the Fig. 2.5, where randomly
generated unit quaternion data set is depicted. The data set is concentrated around
5 clusters each containing 1000 elements generated randomly from Gaussian
2.5 Quaternions Random Generation 15

Fig. 2.4 Visualization of

probability distribution
function for quaternions [66]

Fig. 2.5 Visualization of a

randomly generate
quaternions test set Q 5C

distribution of variance 0.005 with defined quaternion mean. The clusters centers
set Q 5C (in Euler angles form) is defined as:

Q 5C = {(0, 0, 30); (0, 0, 60); (−30, 0, 60); (30, 0, 60); (0, 0, 90); } (2.21)

In case of some further presented algorithms there is also a purpose of the exper-
iment against a different degree of randomly generated data overlapping. As an
overlapping degree we will consider the distance between two adjacent clusters cen-
ters on the unit hypersphere. For example, in the data randomly generated around
two clusters, the sample procedure of changing the amount of overlapping in the data
set can be defined as following:
Q 10dist = {(0, 0, 30); (0, 0, 40)}
Q 20dist = {(0, 0, 30); (0, 0, 50)}
Q 30dist = {(0, 0, 30); (0, 0, 60)}
Q 40dist = {(0, 0, 30); (0, 0, 70)}
Q 50dist = {(0, 0, 30); (0, 0, 80)}
Q 60dist = {(0, 0, 30); (0, 0, 90)}
Q 70dist = {(0, 0, 30); (0, 0, 100)}
16 2 Processes Described by Quaternion Models

Fig. 2.6 Visualization of 3 datasets with variable amount of overlapping clusters. Left - Q 10dist ,
Middle - Q 40dist , Right - Q 70dist

Summarizing, the changing of the data overlapping degree between two(or more)
clusters is a increasing or decreasing of the value of one of its Euler’s angle, which
results in changing the distance between the same rotations represented on the
unit quaternion hypersphere. The example data sets with different data overlapping
amount are presented in the Fig. 2.6

2.6 Advantages and Disadvantages of Using Quaternions

Parametrization of Rotation

All of the reviewed quaternions properties will be used in the proposed algorithms
designed to investigate chaos properties in a quaternion domain. The analysis of
chaos in rotational data will be performed on unit quaternions parametrization due
to their advantages against the other available parametrization.
Closing the chapter about processes described by quaternion models and quater-
nions properties it is worth to recall the main advantages and disadvantages of using
quaternions as a rotation representation.
The main merit of using quaternions parametrization of rotations is they do not
suffer from gimbal lock and coordinates singularities. This property is an outcome
from the Hairy Ball Theorem [124]. The main conclusion from this theorem is that,
no matter how a point is moving continuously around the sphere, there is no such area
where it will be moving in a direction, which is prohibited because of coordinates
singularity presence.
One may say that the representation of a rotation in a quaternion form is consistent
in the sense it requires only four numbers to describe a one rotation and only one
constraint - unit magnitude. This fact implies also a better computational efficiency
and stability than matrix of coordinates. Multiplying quaternions is less complex and
more computationally efficient than multiplying operation on matrices and provides
better numerical stability. It is also worth to underline that even if there occurred
2.6 Advantages and Disadvantages of Using Quaternions Parametrization of Rotation 17

the numerical drift away of unit magnitude, it can be resolved by renormalizing the
given quaternion [65].
In case of disadvantages of using quaternions representation of rotation, the main
one is that they are less intuitive in understanding and usually some additional oper-
ations are required to see the unit quaternions underlings rotation. Moreover, not all
mathematical operations can be implemented in the quaternions world. For example
the Euclidean vector space interpolation methods are not available and there is a need
of having the equivalents in the quaternions domain. One may say the drawback is
also a dual representation of rotation. Two quaternions q and −q represent exactly
same rotation as it was shown in the quaternions averaging section.
Chapter 3
Deterministic Chaos Properties

3.1 Dynamical Systems

The deterministic chaos term and its properties is tightly connected with dynamical
systems theory. Good references of dynamical systems theory can be found in [75,
82, 101, 135, 143].
As it was well summarized in [75] there are a few approaches of modern dynamical
systems theory e.g. Ergodic theory, Topological dynamics, The theory of smooth
dynamical systems and Hamiltonian dynamics.
We assume that in the scope of this book are smooth dynamical systems. As it
was already mentioned in the previous chapter the dynamical system includes three
components:
• phase space
• time
• time evolution law
We assume the phase space represents all possible states of the system. Let’s assume
it is a compact Riemannian manifold M , a compact manifold with Riemannian metric
and sufficiently differentiable structure, which can be rewritten as bounded subset in
Rn . A state of the system at a given time, denoted as X (t), we call the single point in
the manifold M  
X (t) = x(1) , x(2) , . . . , x(d ) (3.1)

where X (t) ∈ M .
By the time we mean a real value (t ∈ R) for the continuous time dynamical
systems or an integer (t ∈ Z) for discrete time systems. The time evolution law
describes how a particular point in manifold M moves in this manifold with time
passing.
The time evolution law is a rule, which describes where the system state is in M
after time t. This rule can be described e.g. by a flows or maps [33].

© Springer International Publishing AG, part of Springer Nature 2019 19

M. Piórek, Analysis of Chaotic Behavior in Non-linear Dynamical Systems,
Studies in Systems, Decision and Control 160,
https://doi.org/10.1007/978-3-319-94887-4_3
20 3 Deterministic Chaos Properties

A continuous time dynamical system is called flow. To describe it’s evolution in

time usually a differential equations are used as it was exemplified in the previous
chapter. Equation 2.1 can be rewritten in more general form as:

dX (t)
= F(X (t)) (3.2)
dt
where X (t) is a state vector and F : M → M is a vector field differentiable as often
as required.
Having the vector field F defined in Eq. 3.2 one may see time variable t does not
appear explicitly. This kind of systems not explicitly dependent on time are called
autonomous and they are the subject of this research. The analysis and embedding
of forced systems is also a subject of the related work e.g. [128–130].
Dynamical evolution of the system is an initial value problem, which resolution
determine what will happen with the initial system’s state X (0) after time t

F : X (0) → X (X (0), t) (3.3)

Having the time evolution law F and initial conditions: X (0), one may resolve
the state of the system in the future time t > 0

X (t) = F(X (0), t) (3.4)

During the evolution, the state of the system draws a path in the phase space called
trajectory or orbit. An example trajectory is presented in the Fig. 3.1.
For the autonomous dynamical system the solution of the initial value problem
of Eq. 3.3 exists and is unique once the vector field F is Lipshitz continuous [75].
The Lipschitz continuity can be defined as following. Let U be an open set in n .
A vector field F(X , t) on n is is said to be Lipschitz on U if does exist a constant
L such that
F(X1 , t) − F(X2 , t) ≤ L X1 − X2  (3.5)

Fig. 3.1 The sample

dynamical system trajectory
3.1 Dynamical Systems 21

where X1 , X2 ∈ U . The constant L is called a Lipschitz constant for F. If the constant

L does exist the vector field F is called Lipschitz-continuous on U [4]. We assume
in the scope of this work are dynamical systems described by Lipschitz vector fields
and Lipschitz maps.
In the case of discrete time (valued by integers), systems are called maps and the
evolution law is given by a map M (Xn ), which maps vectors in M to another vectors
in M :
Xn+1 = M (Xn ) (3.6)

where n stands for the discrete time moments (0, 1, 2, . . .) and

Xn = (xn(1) , xn(2) , . . . , xn(d ) ) (3.7)

is a state vector.
The behavior of the trajectory of dynamical system’s evolution depends both on
the form of F or M and on the initial condition.
It is worth to emphasize, that there exist some sets of initial conditions, which
leads to the same asymptotic behavior of trajectories. Such kind of a set of initial
conditions is called the basin of attraction. The attracting subset of the phase space
to which the trajectories tend to is called an attractor [72]. The examples of attractors
are fixed point attractor and limited cycle attractor presented in the Fig. 3.2.
The dynamical systems can be divided into two subgroups:
• Linear dynamical systems
• Nonlinear dynamical systems
Linear dynamical systems satisfy the superposition principle and they can be
described by the equations using linear operators. Nonlinear dynamical systems do
not satisfy the superposition principle and to describe them nonlinear operators are
used.
It is worth to underline that in practical applications very often the continuous time
dynamical system can be replaced by discrete time system by sampling in discrete
moments in time. It can be also done using Poincare section method ([96]), which
reduces the dimensionality of the system from N to N − 1.

Fig. 3.2 Example of dynamical systems attractors. Left - fixed point attractor, Right - limited cycle
attractor
Another random document with
no related content on Scribd:
alli no se la hauia dado, la qual
dezia en esta manera.

CARTA DE FLAMIANO Á
VASQUIRAN
Verdaderamente, Vasquiran, tus
cartas me desatinan porque
quando miro en ellas el
encarecimiento de tu daño me
parece grande, quando considero
la causa dél lo juzgo pequeño.
Pero en esta carta tuya postrera
he conocido en las cosas que me
escribes lo que te engañas, en
especial en quererte hazer ygual
en el martirio con Petrarca y
Garcisanchez. Si supiesses de
quan lexos vas errado,
maravillarte yas por cierto. Los
tiros de su combate muy lexos
hizieron los golpes de donde los
tuyos dan. De virgines y martires
ganaron ellos la palma si bien lo
miras, que no de confessores de
sus vitorias como tú hazes. Si
gozo ellos han hauido, en la
muerte lo habrian; que en la vida
nunca lo houieron. Mi dolor
sintieron e tu gozo ignoraron.
Claro está segun muestran las
liciones del uno e los sonetos del
otro, e quanto ambos escriuieron,
porque de ninguno dellos leemos
sino pesares en la vida, congoxas
y dolores en la muerte; desseos,
sospiros, ansias apassionadas,
cuydados e disfauores e
desesperados pensamientos;
quando quexando, quando
plañendo, quando pidiendo la
muerte, quando aborreciendo la
vida. Destos misterios dexaron
llenos de tinta sus papeles e de
lastimas su memoria, estos
hizieron sus vidas llenos de pena
e sus fines tan doloridos; con
estos que son los males do mis
males se engendran, con estos
que fueron martirizados como yo
lo soy; verdad es que de dias
vencieron como tú a quien de
amor y fe vencidos los tuvo e los
hizo viuir desseando la muerte
con mas razon que tú la desseas.
Assi que mira lo que por la boca
escriuiendo publicaron e
conoceras lo que en el alma
callando encubierto suffrieron, e
mira si hallarás en ellos vn dia de
victoria como tú plañes doze años
de gloria que dizes que perdiste.
Yo digo que los ganaste, mas
hate parecido a ti que la fortuna te
era obligada a tenerte queda la
rueda en la cumbre del plazer; yo
te prometo que si de sus bienes
no te houiera hecho tan contento,
que de sus males no fueras tan
quexoso sin razon, como estos e
yo lo somos. Tambien me
escriues como soñaste que viste
en vision tu alegria, tus placeres,
tu descanso, tu consentimiento, tu
esperança, tu memoria, tu
desseo; beato tú que primero las
gozaste en la vida y en la muerte
las ensueñas, yo te prometo que
avnque mi placer, ni mi alegria, ni
mi descanso, ni mi
contentamiento, ni mi esperança
yo los encontrasse a medio dia,
que no los conociesse pues que
nunca los vi; mi desseo y mi
memoria no me los cale soñar,
que velando me hazen soñar la
muerte sin dormir cada hora.
Tambien me escribes que viste á
Violina e te habló, e quexaste
dello, ¿qué te pudo hazer
viuiendo que muerta no te quiere
oluidar? No me alegraré yo de lo
que tú, que ni agora en vida ni
despues de mis dias acabados de
mi tuuo memoria ni terná, no digo
de verme que es impossible, mas
avn de pensar si soy en el mundo.
Contentate pues, recobra tu
juyzio, no des mas causa para
que las gentes te juzguen, no
corrompas la reputacion de tu
fama, ni el agudeza de tu ingenio
con tan flaca causa, dando lugar
a tu dolor que de pesar te haya de
tener tal que á ti pierdas e a mi no
ayudes, pues que vees que mi
vida penando se consume; sino te
voy a ver es por la necesidad que
tengo que a verme vengas. Lo
qual te pido que hagas tanto
caramente quanto rogartelo
puedo, porque avnque soledad
busques para tu descanso, la
compañia de mis sospiros te la
dará, e con la mucha confianza
que de ti tengo quedo con tu vista
esperando la respuesta glosando
esta cancion:

Sin remedio es mi herida

pues se cansa quando os veo
y en ausencia mi desseo
más dolor me da en la vida.
¿Qué remedio haurá en mi
pena
si veros fue causa della
y el dolor de mi querella
vuestra ausencia lo condena?
de suerte que no hay salida
para mi, ni yo la veo,
pues veros é mi desseo
son el cabo de mi vida.

LO QUE VASQUIRAN ORDENÓ

DESPUES DE LEYDA LA
CARTA, E COMO SE PARTIO
PARA NOPLESANO.
Otro dia Vasquiran despues de
leyda la carta de Flamiano, de
gran mañana se fue a caça de
ribera y lleuó a Felisel consigo, al
qual despues de hauer volado
una pieça del dia le dixo
tomandole aparte: Ya sabes,
Felisel, como tengo deliberado de
yr a ver a tu señor, porque pues
mis congoxas no bastan para
acabarme quiças las suyas lo
haran; quissiera tenerte comigo
para lleuarte por el camino para
mi descanso e no es cosa que
hazerse pueda por la necesidad
que Flamiano tiene de ti, en
especial con mi yda e tambien
porque no seria razon tomalle
impensado, assi que más eres
allá menester para seruir a
Flamiano que no acá para mi
plazer pues no le tengo, assi que
mañana te parte y darle has
aviso, e pues que yo allá sere tan
en breue, no le delibero escriuir
sino que solamente de mi parte le
digas que si su señora le ha
mostrado sospirar que consigo
aprendera bien á llorar; e assi
hablando se tornaron a Felernisa.
Otro dia Felisel se partió e llegado
que fue á Noplesano fizo saber a
Flamiano la venida de Vasquiran.
Sabido que Flamiano la houo
mandó aparejar dentro en su
posada vn aposento para
Vasquiran, el qual se contenia con
vn jardin que en la casa hauia el
qual mandó adereçar conforme a
la voluutad e vida del que en el
hauia de posar.

LO QUE VASQUIRAN HIZO

DESPUES DE PARTIDO
 FELISEL HASTA LLEGAR A
NOPLESANO
Partido Felisel, Vasquiran deliberó
de yr aquel camino por mar e
mandó fletar vna muy buena naue
de las que en el puerto hauia, e
mandó meter en ella las cosas
que hauia necessarias para el
camino, y embarcar la ropa e
caualgaduras que deliberaua
lleuar; e assi partia á su heredad
ante de embarcar por visitar la
sepultura de Violina. Llegado alli
vna tarde mandó sobre la tumba
pussiesen un titulo con esta letra:

Aqui yaze
todo el bien que mal me haze.

E assi mandó dar orden en todo

lo que en ausencia suya deuia
hazer assi en el concierto de la
casa como en los officios de la
capilla, e assi despidiendose a la
partida hizo esta cancion a la
sepultura:

Pues mi desastrada suerte

contigo no me consiente,
quiero ver si estando ausente
pudiesse hallar la muerte.
Lo que mi viuir querria
es no verse ya comigo
porque yo estando contigo
más contento viuiria,
e pues que veo qu'en verte
 mi pena descanso siente,
cierto so que estando ausente
no verna buscar la muerte.

Otro dia se tornó a Felernisa e

queriendo partirse para
Noplesano mandó poner sobre el
portal de su casa un titulo que
dezia:

Queda cerrada la puerta

que la muerte halló abierta.

Aquesta noche mandaron

embarcar sus servidores, él se
embarcó ante que fuesse de dia
por escusarse de la importunidad
de las visitaciones e de los que al
embarcar le houieran querido
acompañar, hauiendo empero
visitado algunas personas
principales a quien la raçon e
alguna obligacion le constriñia.
Pues siendo ya embarcado
queriendo la naue hazer vela ante
que amaneciese, hizo esta
cancion:

El morir vino a buscarme

para matar mi alegria,
e agora que yo querria
no me quiere por matarme.
El me vino a mi a buscar
teniendole aborrecido
e agora que yo le pido
no le halla mi pesar,
 assi que haurá de forçarme
a buscalle mi porfia
pues veo que se desuia
de mi para más matarme.

Hecho que houo vela la naue, en

pocos dias fueron a vista de la
tierra de Noplesano, e por hauer
tenido algo el viento contrario
hallaronse algo baxos del puerto,
e no podiendole tomar acordaron
por aquella noche de surgir en
vna costa que está baxo de dicho
puerto a quarenta millas de
Noplesano, la qual es tan aspera
de rocas e peñas e alta montaña
que por muy pocas partes se
puede andar por ella a cauallo,
empero es muy poblada de
jardines e arboles de diuersas
maneras, en especial de
torongeros e sidras e limones e
toda diuersidad de rosas, e
muchas caserias assentadas por
lo alto de las rocas; e a la marina
hay algunos lugares e vna gentil
cibdad que ha nombre Malhaze
de donde toma el nombre la
costa. Pues assi llegados, la naue
surgió en vn reparo del viento que
venian muy cerca de tierra, en el
qual lugar, ya otra vez hauia
estado Vasquiran trayendo
consigo a Violina hauia mucho
tiempo. Pensar se puede lo que
Vasquiran sentiria viniendole a la
memoria, la qual le renouo
infinitos e tristes pensamientos los
quales le sacauan del coraçon
entrañables sospiros e infinitas
lagrimas, las quales porque mejor
e mas encobierto derramallas
podiesse, con una viuela en la
mano, de la nao se salio e
sentado sobre una roca muy alta
que la mar la batia, debaxo de vn
arbol començo a cantar esta
cancion:

No tardará la vitoria
de mi morir en llegar,
pues que yo vi este lugar
qu'era tan lleno de gloria
quanto agora de pesar.
Yo vi en toda esta riuera
mill arboles de alegria,
veola agora vazia
de plazer de tal manera
que me da la fantasia
qu'el dolor de su memoria
ya no dexará tardar
mi morir de no llegar
para darme tanta gloria
quanto m'a dado pesar.

Estando alli assi cantando e

pensando acordose que en aquel
mismo lugar hauia estado,
quando por alli passaron él e
Violina e otras señoras que en la
naue venian, toda vna tarde a la
sombra de aquel arbol jugando a
cartas e razonando, e hauian
cenado con mucho plazer
mirando la mar, e assi
acordandose dello començo a
cantar este villancico.

Di, lugar sin alegria,

¿quién te ha hecho sin plazer
que tú alegre solias ser?
¿Quién ha hecho tus
verdores
e tus rosas e tus flores
boluer todas en dolores
de pesares e tristuras,
quién assi t'a hecho ascuras
tus lumbres escurecer
que tú alegre solias ser?

Passada parte de la noche, ya

Vasquiran recogido en la naue,
con el viento de la tierra hizieron
vela e llegaron a hora de missa al
puerto de Noplesano. Mandó
Vasquiran que ninguna señal de
alegria la naue en la entrada
hiziesse de las que acostumbran
hazer. Sabido Flamiano por un
paje suyo que de unos corredores
de su casa vio la naue entrar, lo
que en la entrada hauia hecho,
penso lo que podía ser, e con
algunos caualleros mancebos que
con él se hallaron, sin más
esperar junto con ellos al puerto
se vino, e llegaron al tiempo que
la naue acabaua de surgir, e assi
todos apeados en vna barca en
ella entraron e hallaron a
Vasquiran que se queria
desembarcar. E assi se recibieron
con mucho amor e poca alegria.
Estando assi todos juntos
teniendo Flamiano a Vasquiran
abraçado, en nombre de todos
ellos le dixo: Vasquiran, a todos
estos caualleros amigos tuyos e
señores e hermanos mios que
aqui vienen o son venidos a verte,
no les duele menos tu pesar que
a mi; con tu vista se alegran tanto
como yo. Al qual él respondió:
Plega a Dios que a ti e a ellos
haga tan contentos con la vida,
como a mi con la muerte me
fazia. Al qual respondio el
marques Carlerin: Señor
Vasquiran, para las aduersidades
estremó Dios los animos de los
caualleros como vos, pues que no
es menos esfuerzo saber suffrir
cuerdamente que osar venzer
animosamente. Vasquiran le
respondio: Verdad es, señor
marques, lo que dezis, pero
tambien hizo Dios a los discretos
para saber sentir las perdidas,
como a los esforçados para
gozarse de las ganancias de las
vitorias, e no es menos virtuoso el
buen conocimiento que el buen
animo, ni vale menos la virtud por
saber bien doler, que saber bien
sofrir e osar bien resistir.
E assi razonando en muchas
otras cosas semejantes, salieron
de la naue, e todos juntos vinieron
a la posada de Flamiano donde
hallaron muchos caualleros que
los esperauan, e todos juntos alli
comieron hablando de muchas
cosas. E assi aquel dia passaron
en visitas de los que a ver
vinieron a Vasquiran y de muchos
señores que a visitar le embiaron.

LO QUE VASQUIRAN HIZO

DESPUES DE LLEGADO
Á NOPLESANO
Otro dia despues de hauer
comido, Vasquiran acordo de yr a
besar las manos a la señora
duquesa de Meliano e a Belisena,
e despues al visorey e al cardenal
de Brujas e a la señora princesa
de Salusana e a algunas otras
personas que sus estados e la
raçon lo requeria. E assi
acompañado de algunos
mancebos que con él e con
Flamiano se hallaron, hauiendolo
hecho saber a la señora duquesa
se fueron a su posada, y yendo
por el camino, Flamiano se llegó a
Vasquiran e le dixo: agora ymos
en lugar donde tú de tus males
serás consolado e yo de los mios
lastimado. Al qual respondio
Vasquiran: mas voy a oyr de
nueuo mis lastimas; tu vás a ver
lo que desseas; yo recibire pena
en lo que oyre; tú recibiras gloria
en lo que verás. Assi razonando
llegaron a la posada de la señora
duquesa, a la qual hallaron en
vna quadra con aquel atauio que
a tan gran señora siendo uiuda se
requeria, acompañada de la
señora Belisena su hija, con todas
las otras damas e dueñas de su
casa. E como las congoxas de los
lastimados con ver otros llagados
de su herida no pueden menos de
no alterar el dolor de las llagas,
alli hauiendo sido esta noble
señora vna de las que con más
raçon de la aduersa fortuna
quexarse deuia, uiendole perder
en poco tiempo el catolico abuelo,
la magestad del serenissimo
padre, el clarissimo hermano en
medio del triunfo mas prospero de
su gobierno reynando, e sobre
todo el ylustrissimo marido tan
tiranamente de su estado e
libertad con el heredero hijo
desposseidos, de manera que no
pudo menos la vista de Vasquiran
hazer que de mucho dolor su
memoria no lastimasse, e
verdaderamente ninguna de las
que viuen para ello mas raçon
tiene.
Pues assi llegados, hauiendo
Vasquiran besado las manos a la
señora duquesa, e a Belisena
hecho aquel acatamiento que se
deue hazer e a todas las otras
señoras e damas, despues de
todos sentados, la duquesa
començó de hablar en esta
manera.

LO QUE LA SEÑORA DUQUESA

HABLÓ A VASQUIRAN EN
PRESENCIA DE TODOS; E
LO QUE VASQUIRAN LE
RESPONDIO E ALLI PASSÓ.
Vasquiran, por vida de mi hija
Belisena qu'es la mas cara cosa
que la fortuna para mi consuelo
me ha dexado, que considerado
el valor e virtud e criança tuya, y
el amor e voluntad que al duque
mi señor, que haya santa gloria, e
a mi casa siempre te conoci tener,
sabido tu perdida tanto tu daño
me ha pessado, que con los mios
ygualmente me ha dado fatiga.
Esto te digo porque conozcas la
voluntad que te tengo, lo que
consolarte podria remitolo a ti
pues te sobra tanta discrecion
para ello quanto a mí me falta
consuelo para mis males.
Vasquiran le respondio: Harto,
señora, es grande mi desuentura
quando en tan alto lugar ha hecho
señal de compasion, mas yo doy
gracias a Dios que me ha hecho
tanto bien en satisffacion de tanto
mal qu'en tan noble señora como
vos e de tan agrauiados males
combatida mi daño haya tenido
cabida o lugar de doler; lo que yo
señora siempre desseo vuestro
seruicio Dios lo sabe; lo que en
vuestras perdidas yo he sentido
ha sido tanto que el dolor dellas
tenia ya en mí hecho el aposento
para quando las mias llegaron.
En esto y en otras cosas
hablando llegó el tiempo de
despedirse, en el que nunca
Flamiano los ojos apartó de
Belisena. Pues siendo de pies ya
de la duquesa despedidos,
Vasquiran se despidio de
Belisena a la qual dixo: señora,
Dios os haga tan contenta como
vos mereceys e yo desseo,
porque ensanche el mundo para
que sea vuestro y en que mi
pesar pueda caber. Al qual ella
respondio: Vasquiran, Dios os dé
aquel consuelo que con la vida se
puede alcançar, de manera que
tan alegre como agora triste
podays viuir muchos dias. E assi
la señora Yssiana se llegó a ellos
e muy baxo le dixo: señor
Vasquiran, esforçaos, que no
juzgo menos discrecion en
vuestro seso que dolor en vuestro
pesar; la fortuna os quitó lo que
pudo, pero no la virtud que en vos
queda que es más.
Señora, dixo Vasquiran, plega á
Dios que tanta parte os dé la
tierra quanta en vuestra
hermosura nos ha dado de lo del
cielo, pues que está en vos mejor
aparejado el merecer para ello
que en mí el consuelo para ser
alegre. Bien sé yo que si posible
fuera que en mí pudiera hauer
remedio para mi tristeça, el
esperança de vos sola la
esperara.
Al qual respondio la señora
Persiana: Vasquiran, por la
compasion que tengo de ver
vuestra tristeça, quiero consentir
que me siruays e sin perjuizio mio
yo hare que perdays mucha parte
de vuestra passion con mis
fauores.
Assi tornado a la señora duquesa
se despidio con todos aquellos
caualleros que con él hauian
venido, e quedose alli el marques
Carlerin. De alli se fueron a visitar
al señor visorey con el que
hallaron al cardenal de Brujas y el
cardenal de Felernisa, los quales
todos con mucho amor le
recibieron. El restante de lo que
alli passó, por abreuiar aqui se
acorta. Assi se tornaron á su
posada. Otro dia fue a besar las
manos a la reina Noplesana e a
su madre, e despues a otras
muchas señoras que a la sazon
en Noplesano se hallaron.

LO QUE DESPUES DE LAS

VISITACIONES E HAUER
REPOSADO ALGUNOS DIAS,
ENTRE FLAMIANO Y
VASQUIRAN PASSÓ SOBRE
SU QUESTION
Estando vn dia acabado de comer
Vasquiran e Flamiano en vna
huerta de su posada acostados
de costado sobre vna alfombra
debaxo vnos naranjos, comemço
Vasquiran en esta manera de
dezir. No quiero, Flamiano, qu'el
plazer de nuestra visita con su
plazer ponga silencio en nuestra
question a sus pesares, porque
tanto por dalle fin a nuestra
question soy venido, quanto por
verte; a tu postrera carta no
respondi por hazerlo agora.
Muchas variedades he visto en
tus respuestas assi de lo que en
mi contradizes como de lo que en
ti manifiestas, en especial agora
que a Belisena he visto, e digo
que todo el fin de tu mal seria
perder la vida por sus ameres;
digote vna cosa, que si tal
perdiesses el más de los bien
auenturados te podrias llamar,
¿pues si tu muerte seria
venturosa, tu pena no es
gloriosa? claro está. Todas las
cosas que me has escripto en
cuenta de tus quexas, agora que
lo he visto juzgo en cuenta de tus
glorias; quando nunca más bien
tuuiesses de verte su servidor es
mucho para hacerte ufano,
quanto más que tus ojos la
pueden ver muchas veces, que
más bien no le hay. Quantas
cosas me podrias encarecer de
los males que pregonas no son
nada, por que Quiral en tu egloga
te ha respondido lo que yo podria;
digote vna cosa, que te juzgo por
mas dichoso penando en seruicio
suye que no si alegre te viese sin
seruilla. Si assi supiesses tú suffrir
contento tu pena como supiste
escoger la causa della, ni comigo
competerias como hazes, ni yo te
reuocaria como hago. No plega a
Dios que mi mal sepas a qué
sabe, ni de tu pena sanes porque
viuas bien auenturado. Mirado el
lugar do tu desseo e voluntad
possiste, de todo lo possible
gozas; visto lo que quexas, todo
lo impossible desseas. Visto lo
que yo perdi no hay mas bien que
perder; visto lo que yo desseo no
hay mas mal que dessear, pues
que al fin con la vida se acaba
todo.
A todas las cosas que me has
escripto te he respondido; a lo
que agora me querras dezir
tambien lo verás, oyrte quiero.
 RESPUESTA DE FLAMIANO
Vasquiran, todo quanto hasta
agora en mis cartas y de palabra
te he escripto y enbiado a dezir,
en dos cosas me parece que
consiste. La vna, ha sido
parecerme que quexas mas de lo
que deues e que no perdiste sino
que se acabó tu plazer, e que
demasiado estremo dello
muestras. La otra ha sido que mi
mal es mayor qu'el tuyo. Agora
quiero que despacio juntos lo
determinemos, e quiero començar
por mí. Dizesme que las virtudes
e merecimientos de Belisena con
quantas excelencias en ella has
visto, me deuen hazer ufano y
contento, e que si por ella
perdiesse la vida seria bien
auenturado, e que no puedo mas
perder, e que cada hora la veo,
que no hay más bien que perder e
que desseo lo impossible y gozo
lo possible. ¿Cómo se podra
hazer que las perficiones de
Belisena si estas mismas
encienden el fuego do m'abraso
hagan mi pena gloriosa? quanto
más de su valer contento, más de
mi remedio desconfio, e si como
dizes por ella la vida perdiesse,
bien dizes que seria bien
auenturado, mas no la pierdo y
muero mill vezes cada hora sin
que agradecido me sea; el bien
que me cuentas que por su vida
gano, es todo el mal que cada
hora renueua mis males, pues
que para más no la veo de para
mis pesares. Pues mi desseo es
impossible, ¿qué bien puedo
hauer que sea lo posible como tú
dizes? A mi me pareze que el fin
de todas las glorias está en
alcançarse e no en dessearse,
porque el desseo es un acidente
que trae congoxa, e quanto mayor
es la cosa deseada mayor es la
congoxa que da su desseo;
¿pues cómo me cuentas tu a mi
el desseo por gloria siendo él
mismo la pena? Visto estar claro
que de todas las cosas e desseos
se espera algun fin, de todos los
trabajos se espera algun
descanso. Todos los desseos se
fundan sobre alguna esperança,
porque si cada cosa destas esta
causa no la caussase, no ternia
en si ninguna razon, pues que no
tuuiesse principio donde naciesse
no ternia termino do acabase,
pues no teniendo principio ni cabo
consiguiente caduca seria. Pues
luego si mi desseo es impossible
y es grande y grande la pasion
que me da, ¿qué cuenta haura en
mi mal? no otra sino que no hay
remedio para él? Pues si el
remedio le falta, el mio es grande,
que el tuyo no.