Manfred Constapel, M.Sc.

M. Constapel
Position Former Staff


Since August 2017 I was research associate at the Center of Excellence CoSA at the Lübeck University of Applied Sciences. For my Master thesis I worked in the field of sensor fusion and air surveillance and developed a virtual testbed for evaluation of filters and motion models for radar tracking. I'm interested in software development, embedded systems and data science. Some of my software projects can be found on GitHub. To get an overview about my publications visit SemanticScholar.


1979 born in Norden, Germany
1998 - 2001 Apprenticeship as Specialized Computer Scientist in Application Development
2002 Joining the German Federal Armed Forces (Navy)
2005 - 2008
Study of Computer Science with focus on Business Information at the Stralsund University of Applied Sciences
2010 - 2015 Submarine officer in operations and nautical service of the German Federal Armed Forces
2014 - 2017
Study of Computer Science with focus on Interactive 3D and Software Technology at the Lübeck University of Applied Sciences


DRAISE Wireless, robust, adaptive, industrial Systems
IKARion Intelligent support to improve small group works in online teaching
RoSiE Robust and Safety-relevant Real-Time Localization



Refereed Articles and Book Chapters
[2018] A new localization algorithm based on neural networks (Mathias Pelka, Manfred Constapel, Duc Tu Le Anh, Horst Hellbrück), In Proceedings of the 3rd KuVS/GI Expert Talk on Localization, 2018. [bib] [pdf] [abstract]
Indoor localization plays a major role in a wide range of applications. To determine the location of a tag, localization algorithm is required. In the past, machine learning algorithms were difficult to implement in consumer hardware, but with the advent of tensor processing units, even smartphones are capable to use artificial intelligence to solve complex problems. In this paper, we investigate a machine learning algorithm based on neural networks and compare the result to a linear least squares estimator. We design and evaluate different neural networks. Based on our observation, the neural network delivers poor performance compared to the linear least squares estimator.
[2018] Drahtlose Robuste Adaptive Industrielle Systeme (Manfred Constapel, Leo Krüger, Swen Leugner, Zeynep Vatandas, Koojana Kuladinithi, Andreas Timm-Giel, Horst Hellbrück), In ITG-Fachbericht-Mobilkommunikation VDE VERLAG GmbH, 2018. [bib] [abstract]
Die industrielle Fertigung wandelt sich mit der Digitalisierung. Fertigungsstücke werden individuell durch Industrie 4.0-Technologien; sie sind jederzeit und überall erfassbar. Fertigungsprozesse werden digitalisiert und bilden Struktur und Verhalten einer Fertigungs- und Produktionsstrasse ab, womit sich wichtige Erkenntnisse zum Ablauf gewinnen lassen. Dies setzt eine Vernetzung der Fertigungsstücke voraus, welche die klassische Automatisierung ergänzt. Drahtlose Sensoren sind an Fertigungsstücke gekoppelt und bilden großflächige und dichte Sensornetze. Messgrößen der Sensoren sind beispielsweise Ort, Zeit, Temperatur, Beschleunigung, Geräusch oder Luftfeuchte, welche nach Anforderung der Datenanalyse ausgewählt und ausgewertet werden. Derartige Sensornetze stellen neue Herausforderungen und Forschungsfragen an Industrie 4.0-Anwendungen. Das Projekt DRAISE hat das Ziel, produktionsrelevante Daten in Industrieanlagen zuverlässig zu erfassen und entsprechend aufzubereiten. Hierfür werden Transportwagen einer Fertigung drahtlos vernetzt und Temperatur, Erschütterung, Position und Zeit erfasst. Zusätzlich interagieren die Mitarbeiter in der Produktion über eine definierte Mensch-Maschine-Schnittstelle mit dem System. Zur Minimierung der Auswirkungen durch andere, drahtlose Systeme werden Time Division Multiple Access (TDMA)-basierende Protokolle für die Single-Hop- und Multi-Hop-Kommunikation entwickelt. Die Kommunikation wird durch Lokalisierungskomponenten ergänzt, welche mit einem System auf Basis von Ultra-Wide-Band (UWB) implementiert ist. Für die angestrebte, zuverlässige und robuste Kommunikation werden zukünftig Frequenzsprungverfahren und Spectrum Sensing die entwickelten Protokolle erweitern. Ein großflächig und produktiv eingesetzter Demonstrator dient der ständigen Leistungsbewertung des Systems.
Refereed Conference Papers
[2019] IKARion: Enhancing a Learning Platform with Intelligent Feedback to Improve Team Collaboration and Interaction in Small Groups (Manfred Constapel, Dorian Doberstein, H. Ulrich Hoppe, Horst Hellbrück), In Proceedings of the 18th International Conference on Information Technology Based Higher Education and Training, 2019. [bib] [abstract]
Due to a high number of participants and limited human tutoring capacities, close supervision and human guidance is hardly possible in large online courses. Thus, the aim of the IKARion project is to develop methods for intelligent and automated diagnosis and intervention in online learning environments. In this paper, we present our distributed and autonomous feedback system aiming to improve team collaboration and interaction by means of interventions. The feedback system detects and diagnoses existing and emerging problems within the groups based on their interaction patterns. Based on principles of pedagogy and educational psychology, analysis metrics and appropriate measures are provided for different types of feedback including mirroring for exploratory and informative prompts providing appropriate information about progress and actual state for the teamwork, and guiding prompts aiming to modify the interaction and collaboration of a team or an individual student. The distributed feedback system consists of three essential components: Moodle-based learning management system along with various prompts implemented as Moodle plugins, Learning Analytics Backend, and Rule-based Intervention System based on an expert system. We describe, on an architectural level, the distributed feedback system and its components. As a result, we show measurements of two long term experiments carried out in two online courses at university level with a duration of one semester each. We conclude the paper with an outlook on future work.
[2019] Analysis of Types, Positioning and Appearance of Visualizations in Online Teaching Environments to Improve Learning Experiences (J. Brandenburger, M. Constapel, H. Hellbrück, M. Janneck), In International Conference on Applied Human Factors and Ergonomics, 2019. [bib] [abstract]
In this paper we investigate different visualizations of learners’ data related to collaborative online learning in terms of suitability and attractiveness to students. Furthermore, we analyze whether positioning and color appearance of these data visualizations might have an effect on learners’ behavior. To that end, we conducted an online study (n=120) as well as an eye tracking study (n=20) to compare different types of visualizations. Results show that students prefer classical data visualizations like bar charts. Visualizations placed in the sidebar of a two column web interface get less attention than visualizations in the header of the main content area. Color schemes do not seem to influence the perception of visualizations. We discuss possible explanations and implications for designing data visualizations in learning environments.
[2018] TriClock – Clock Synchronization compensating Drift, Offset and Propagation Delay (Swen Leugner, Manfred Constapel, Horst Hellbrück), In IEEE International Conference on Communications, 2018. [bib] [abstract]
In wireless sensor networks (WSN) precise clock synchronization is still a challenge e.g. for synchronized medium access control (MAC). State of the art solutions require many messages or neglect clock drifts or propagation delay. In multi-hop networks synchronization errors increase with the number of hops because numerous messages increase latency. The latency and clock drift reduces synchronization accuracy. Finally, propagation delay introduces additional synchronization offsets. We introduce a novel synchronization protocol that requires a single message to compensate both clock offset and clock drift and one additional message to account for propagation delay. With this minimal amount of messages, an efficient multi-hop synchronization is practicable. We implement our approach on a DWM1000 hardware and evaluate the protocol in single-hop and multi-hop configuration. In our preliminary experiments, we achieved a synchronization accuracy of 0.46 ns in a single-hop configuration within 3.6 ms and 6 ns in a multi-hop configuration for 5 hops within 11 ms which is appropriate for MAC and time-division multiple access (TDMA) implementations.
Other Publications
[2019] A Practical Toolbox for Getting Started with mmWave FMCW Radar Sensors (Manfred Constapel, Marco Cimdins, Horst Hellbrück), Technical report, Technische Universität Braunschweig, 2019. [bib] [pdf] [abstract]
In this paper, we sum up our experience gathered working with mmWave FMCW radar sensors for localization problems. We give a glimpse of the foundations of radar that is necessary to understand the benefit and advantages of this technology. Moreover, we introduce our open-source software toolbox pymmw based on Python for Texas Instruments IWR1443 ES2.0 EVM sensors to provide students and researchers easy access to those radar sensors. In doing so, one can jump right into sensing with mmWave FMCW radar from a practical point of view and start doing experiments and developing own applications. Finally, pymmw is used for data acquisition of a scene illuminated by three virtual radars in three different states of occupancy showing the potential of mmWave FMCW radar for indoor and distance-based localization applications.
[2019] Recurrent Neural Network for Sequence Classification of mmWave Radar Data (T. Geiger, M. Constapel, H. Hellbrück), (T. M. Buzug et. al., ed.), 2019. [bib] [abstract]
This paper unveils the potential and utilization of Recurrent Neural Network (RNN) in radar applications for sequence classification. We focus on human activity recognition, which inputs are multichannel time series signals acquired by a frequency modulated continuous wave (FMCW) millimeter wave (mmWave) radar. We designed a proof of concept implementation for a system where a mmWave radar records the movement of a human body. The proposed RNN predicts and classifies the direction the human is heading. Such systems are valuable in numerous applications ranging from security and safety applications like detection and assessment of human activity at airports to patient monitoring in hospitals. Applied to the dataset recorded by the mmWave radar the implemented RNN provides satisfying classification results.
[2017] A syntactic approach to wreck pattern recognition in sonar images (M. Constapel, T. Teubler, H. Hellbrück), GRIN (T. M. Buzug et. al., ed.), 2017. [bib]
Powered by bibtexbrowser

Supervised Students

[2019] Tim Geiger, Entwicklung, Implementierung und Evaluation von neuronalen Netzen zur Modellierung einer in situ Schmelzbadüberwachung für den SLM-Prozess, Master Thesis.
[2019] Huicheng Qian, Use Case Driven Chirp and Frame Configuration for FMCW Radars, Bachelor Thesis.
[2018] Tim Geiger, Recurrent Neural Network for Sequence Classification of mmWave Radar Data, Project Thesis.
[2018] Sébastien Samson, Test and evaluation of the Cooja Network Simulator, Project Thesis.
[2018] Yuefeng Liu, Modeling, implementation and evalutation of SLAM for determination of anchor configurations, Bachelor Thesis.
[2018] Wiland Arlt, Entwurf, Implementierung und Test einer Software-Lösung zur Ansteuerung eines FTDI-Chips, Bachelor Thesis.
[2018] n.n. , Vergleich von Internet of Things Plattformen und Implementierung einer Smart Home Anwendung, Master Thesis.