Organic Computing Doctoral Dissertation Colloquium 2018 B
Total Page:16
File Type:pdf, Size:1020Kb
13 13 S. Tomforde | B. Sick (Eds.) Organic Computing Doctoral Dissertation Colloquium 2018 Organic Computing B. Sick (Eds.) B. | S. Tomforde S. Tomforde ISBN 978-3-7376-0696-7 9 783737 606967 V`:%$V$VGVJ 0QJ `Q`8 `8 V`J.:`R 1H@5 J10V`1 ? :VC Organic Computing Doctoral Dissertation Colloquium 2018 S. Tomforde, B. Sick (Editors) !! ! "#$%&'&$'$(&)(#(&$* + "#$%&'&$'$(&)(#$&,*&+ - .! ! /)!/#0// 12#$%'$'$()(#$, 23 & ! )))0&,)(#$' '0)/#5 6 7851 999!& ! kassel university press 7 6 Preface The Organic Computing Doctoral Dissertation Colloquium (OC-DDC) series is part of the Organic Computing initiative [9, 10] which is steered by a Special Interest Group of the German Society of Computer Science (Gesellschaft fur¨ Informatik e.V.). It provides an environment for PhD students who have their research focus within the broader Organic Computing (OC) community to discuss their PhD project and current trends in the corresponding research areas. This includes, for instance, research in the context of Autonomic Computing [7], ProActive Computing [11], Complex Adaptive Systems [8], Self-Organisation [2], and related topics. The main goal of the colloquium is to foster excellence in OC-related research by providing feedback and advice that are particularly relevant to the students’ studies and ca- reer development. Thereby, participants are involved in all stages of the colloquium organisation and the review process to gain experience. The OC-DDC 2018 took place in Wurzburg,¨ Germany on June 21st to 22nd, 2018. It was organised by the Games Engineering group at the University of Wurzburg¨ in cooperation with the Intelligent Embedded Systems group at University of Kassel. It continued the series of successful OC-DDC events: • The 2013 OC-DDC took place in Augsburg, Germany [16] • The 2014 OC-DDC took place in Kassel, Germany [17] • The 2015 OC-DDC took place in Augsburg, Germany (together with the TSOS Spring School) [18] • The 2016 OC-DDC took place in Duisburg, Germany [19] • The 2017 OC-DDC took place in Bochum, Germany [20] The 2019 edition will be part of the Informatik20191 conference, organised at Uni- versity of Kassel. Seventeen contributions have been accepted—allowing the authors to present their work in Wurzburg.¨ In addition, valuable input has been provided to the OC-DDC by Sylvain Cussat-Blanc (Universite Toulouse) and Sebastian von Mammen (Uni- 1 https://informatik2019.de/ (last access: 29/01/2019) VIII Preface versitat¨ Wurzburg)¨ as invited speakers. We would like to take this opportunity to express our gratitude to all participants and in particular to the invited speakers. This book presents the results of the OC-DDC 2018. Successful participants have been invited to extend their abstracts submitted to the event towards a full book chap- ter by taking reviews and feedback received at the event in Wurzburg¨ into account. The participants prepared an initial extended abstract, helped to perform a sophisti- cated review process, and finally came up with interesting articles summarising their current work in the context of Organic Computing. Hence, the book also gives an overview of corresponding research activities in the field in Germany for the year 2018. The collection of contributions reflects the diversity of the different aspects of Organic Computing. In the following, we outline the contributions contained in this book. Organic Computing postulates to equip technical systems with “life-like” properties. Technically, this means to move traditional design-time decisions to runtime and into the responsibility of systems themselves. As a result, systems have a dramat- ically increased decision freedom that leads to highly autonomous behaviour. The goal of this process is to allow for self-adaptation and self-improvement of system behaviour at runtime [15]. Especially since conditions that occur at runtime can only be anticipated to a certain degree, efficient mechanisms are needed that guide the system’s behaviour even in cases of missing knowledge or uncertain environmental status. Consequently, Organic Computing summarises a variety of aspects and tech- niques that are needed to finally develop such mechanisms [14]. For instance, a major challenge for Organic Computing and related research initiatives is the increasing in- terconnectedness of autonomous systems [13]. As a result, we face open, distributed systems consisting of autonomous elements belonging to different authorities [5, 3]. In particular, autonomous entities as part of large-scale distributed system composi- tions require secure interaction schemes [6]. In conclusion, the term Organic Computing comprises efforts to develop intelligent systems, i.e., to investigate techniques and concepts that allow for a runtime self- adaptation and self-organisation of typically distributed subsystems. Therefore, ap- proaches from different domains are needed: distributed systems (e.g., self-organised communication, election and consensus algorithms), multi-agent systems (e.g., self- organised system organisation, negotiation), machine learning, and systems engi- neering. In the following, we will see that most of these fields are covered by the contributions of this year’s OC-DDC. The first group of papers is dedicated to a system perspective, although already incorporating machine learning techniques. A first contribution by Andre Bauer (University of Wurzburg)¨ deals with the question how future system states in Autonomic/Organic Computing systems can be predicted reliably. The paper is entitled “Challenges of and Approaches for Forecasting in the Context of Autonomic Computing”. Closely related in terms of the underlying techniques is the work by Markus Gorlich¨ (University of Augsburg). He focuses on predicting the next maintenance cycle, Preface IX i.e., the need for human-based exchange of components: “Predictive Maintenance in Complex Systems”. A third paper is presented by Markus Proll¨ (Hochschule Augsburg)—his work is situated in the context of industrial production and also aims at predicting system behaviour. The contribution is entitled “Transparency in Production Processes - A Big Data Analytic System Approach”. To allow for dealing with unanticipated situations and to continuously improve the system behaviour at runtime, machine learning techniques play an important role within the overall concept of an OC system [12]. In this context, Jan Schneegans and Maarten Bieshaar from University of Kassel presents their joint contribution entitled “Smart Device based Initial Movement Detection of Cyclists using Con- volutional Neuronal Networks”. They summarise the key challenges for real-world applicability of neural network-based learning techniques in the context of predicting the movements of vulnerable traffic participants such as cyclists. Transfer learning is another sub-domain in machine learning. The contribution by Simon Stieber (University of Augsburg) makes use of transfer learning concepts in the field of carbon fiber production. His work is entitled “Transfer Learning for Op- timisation of Carbon Fiber Reinforced Polymer Production”. Also in the context of transfer learning, Jens Schreiber (University of Kassel) focuses on the application area of renewable power generation. He presents the concept for his PhD thesis using the title “Transfer Learning in the Field of Renewable Ener- gies: A Transfer Learning Framework Providing Power Forecasts Throughout the Lifecycle of Wind Farms After Initial Connection to the Electrical Grid”. Wenzel Pilar von Pilchau (University of Augsburg) uses the blockchain and its vari- ous applications as target system. He analyses the applicability of machine learning algorithms to tackle challenges in this domain: “Combining Machine Learning and Blockchain: Benefits, Approaches and Challenges”. As eighth contribution, Andreas Margraf (Fraunhofer IGCV, Augsburg) makes use of a particular field in machine learning: evolutionary computation. He also works in an application-driven scenario and entitled his work with “Evolutionary Learning for Processing Pipelines in Image Segmentation and Monitoring Systems”. The Organic Computing initiative has always considered nature and biological pro- cesses as inspiration for mastering complex systems, see e.g. [4]. In this context, the next three contributions follow the path of Organic Computing. Andreas Knote (University of Wurzburg)¨ presents his work on simulating biological cell structures and the underlying biological processes efficiently. His contribution is entitled “Agent-Based Biological Cell Simulations for Self-Organising Develop- mental Processes in Morphogenesis”. The tenth paper is presented by Oliver Kosak (University of Augsburg) and entitled “Multipotent Systems: A New Paradigm for Multi-Robot Applications”. He consid- X Preface ers scenarios, where the cooperation of a potentially large set of autonomous robots is needed to solve a task. Furthermore, Adrian Calma (University of Kassel) resembles the hybrid system structure of many natural systems. He presents his work on “Decision support with Hybrid Intelligence”. The Organic Computing initiative has its roots in the technical computer science domain [1]. Since the very beginning of the initiative, questions of how to realise self-adaptation and self-learning behaviour in computer architecture have been part of the major focus. This year, we have three contributions that continue research in this direction. Norbert Schmitt (University of