TU Darmstadt / ULB / TUprints

Automatic Derivation of Dependency Chains within Systems for Automated Driving via Ontology Based Scenario Representations

Hoßbach, Phillip Maxim (2019)
Automatic Derivation of Dependency Chains within Systems for Automated Driving via Ontology Based Scenario Representations.
Technische Universität Darmstadt
Master Thesis, Primary publication

math_713_18_ext_am_hossbach.pdf - Updated Version
Copyright Information: In Copyright.

Download (5MB) | Preview
Item Type: Master Thesis
Type of entry: Primary publication
Title: Automatic Derivation of Dependency Chains within Systems for Automated Driving via Ontology Based Scenario Representations
Language: English
Referees: Amersbach, Christian ; Darms, Dr. Michael
Date: 15 April 2019
Place of Publication: Darmstadt
Date of oral examination: 30 April 2019

In the development of automated vehicles, the complexity of the underlying, safety-critical systems poses a challenge for today’s engineers. Furthermore, mandatory standards such as ISO 26262 require traceability and assignability of elaborated requirements. In addition, due to the increasing use of agile product development processes, it is necessary to be able to deal with continuous system changes. In this work an approach is presented, which on the one hand aims at supporting design decisions of developers by representing dependencies in a system. On the other hand, an improved documentation and traceability of system requirements is intended. The proposed approach is based on the automatic derivation of so-called dependency chains. For this, ontology based representations of driving scenarios and a system’s architecture modeled by directed acyclic graphs are utilized to enable associations between scenarios and particular system components.

After an introduction to the corresponding state of the art, the fundamentals of ontology engineering and the representation of system architectures are outlined. In addition a consistent terminology for driving scenarios is adopted and the situation awareness and information acquisition within driving tasks is described. Subsequently, the conceptual basis of the approach is presented. Based on the established terms a meta-model is developed, from which three key challenges concerning the intended solution are derived. These challenges are then addressed by the design of partial solutions. Not only the development methodology of the respective ontology is discussed. Also, a dedicated modeling possibility for a system’s architecture, the task-chain-pattern skill graph representation, is elaborated in this context. Moreover, the construct of a chain derivation engine, which constitutes the core concept of this work, is explained. Semantic rules contained in this engine, together with arithmetic functions attached to it, enable the eventual derivation of the intended dependency chains. In order to provide a proof of concept, the developed solution proposals are implemented first separately then collectively. Therein, driving scenarios are narrowed to two elementary scenarios in order to limit the scope of the implemented ontology. Regarding the exemplary system architecture representation, one particular skill is elaborated. Therefore, dedicated examples are provided for the different components of the implementation. In this way, the process is examined in more detail. The examples are then assembled and observed collectively. This illustrates the holistic chain derivation process and achieved analysis functionalities. The implemented software framework is charged with different quantities of variously complex input information and the resulting runtimes of the chain derivation process are interpreted. Hence, its potential and limits for a utilization in the development of fully automated vehicles is evaluated.

URN: urn:nbn:de:tuda-tuprints-87490
Divisions: Study Areas > Study Area Mechatronics
Date Deposited: 23 Jul 2019 12:30
Last Modified: 09 Jul 2020 02:37
URI: https://tuprints.ulb.tu-darmstadt.de/id/eprint/8749
PPN: 451027108
Actions (login required)
View Item View Item