Unsupervised Pattern Discovery in Automotive Time Series

Pattern-based Construction of Representative Driving Cycles

verfasst von: Fabian Kai Dietrich Noering

Verlag: Springer Fachmedien Wiesbaden

Buchreihe : AutoUni – Schriftenreihe

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Inhaltsverzeichnis

Über dieses Buch

In the last decade unsupervised pattern discovery in time series, i.e. the problem of finding recurrent similar subsequences in long multivariate time series without the need of querying subsequences, has earned more and more attention in research and industry. Pattern discovery was already successfully applied to various areas like seismology, medicine, robotics or music. Until now an application to automotive time series has not been investigated. This dissertation fills this desideratum by studying the special characteristics of vehicle sensor logs and proposing an appropriate approach for pattern discovery. To prove the benefit of pattern discovery methods in automotive applications, the algorithm is applied to construct representative driving cycles.

Inhaltsverzeichnis

Frontmatter

Chapter 1. Introduction

Abstract

Unsupervised pattern discovery defines the task of finding similar recurring subsequences in large time series without having any knowledge about the shape, length or level of detail of these patterns. Existing pattern discovery algorithms have already been successfully applied to various research areas like seismology, medicine or music. In seismology they have been used to discover repeating earthquake sequences.

Fabian Kai Dietrich Noering

Chapter 2. Related Work

Abstract

This chapter includes an extensive review of related work in literature. As this work can be separated in two parts, i.e. the general method of unsupervised pattern discovery in time series and its application for representative driving cycles, this chapter will address both topics separately, too. Section 2.1 begins with a discussion of the term pattern, its range and diversity. This is necessary in order to align the expectations and the focus of this work. Beside this, the section gives an introduction to the topic of pattern discovery in time series and gives an overview of existing approaches. Additionally, various challenges and requirements are discussed. Section 2.2 gives an introduction into the field of representative driving cycles including common applications, the definition of representativeness and existing approaches for the construction. Finally, Section 2.3 points out the obvious, but yet not discovered, linkage of both topics.

Fabian Kai Dietrich Noering

Chapter 3. Development of Pattern Discovery Algorithms for Automotive Time Series

Abstract

This chapter deals with the development and optimization of algorithms for unsupervised pattern discovery in time series. While this work focuses on automotive time series data, the proposed approach, in general, is able to improve the handling of highly dynamic sensor data independently of its origin.

Fabian Kai Dietrich Noering

Chapter 4. Pattern-based Representative Cycles

Abstract

The pattern discovery is able to solve a variety of problems related to time series as it is able to adapt to the use cases requirements. To exemplarily demonstrate its benefit, this chapter deals with the idea and implementation of pattern-based RDCs as already outlined in Section 2.3.

Fabian Kai Dietrich Noering

Chapter 5. Evaluation

Abstract

In previous sections algorithms were proposed for unsupervised pattern discovery in multivariate time series and applied for the construction of pattern-based RDCs. This section aims to validate and evaluate the proposed algorithms and approaches. First, Section 5.1 deals with the pattern discovery itself independently of the use case of RDC construction.

Fabian Kai Dietrich Noering

Chapter 6. Conclusion

Abstract

Unsupervised pattern discovery deals with the identification of similar recurring subsequences in time series without any knowledge regarding these patterns. The unsupervised discovery of patterns in time series has proven to be beneficial in many different research areas. This thesis investigated the application of pattern discovery in automotive use cases, which was considered as a gap in research previously. The handling of automotive data, or vehicle time series data, is especially challenging because it unites properties like high dynamism, diversity, dimensionality and large sizes.

Fabian Kai Dietrich Noering

Backmatter

Titel: Unsupervised Pattern Discovery in Automotive Time Series
verfasst von: Fabian Kai Dietrich Noering
Verlag: Springer Fachmedien Wiesbaden
Electronic ISBN: 978-3-658-36336-9
Print ISBN: 978-3-658-36335-2
DOI: https://doi.org/10.1007/978-3-658-36336-9