Photo: Oliver Kussinger

Together with our partners, we are working on many exciting projects involving AI. The most important information about our ongoing and completed projects can be found below. If you have any questions or would like more details, please feel free to contact us!

Ongoing projects

Process Mining: Servo Press

The Rauschert Group, the pioneering manufacturer of technical ceramics, moulded plastic parts, components and assemblies, is the global leader in the development and production of ignition systems for heating systems, such as ignition and ionization electrodes for gas and oil burners. In the production of these ignition systems, a metal retaining plate is joined to the ignition ceramic using a press.

Click here to learn more.

Predictive Maintenance: Belt Grinder

This project investigated the possibility of determining the wear of a grinding belt and the presence of a fault in a belt grinder via acoustic measurements using AI models and mathematical methods. The aim was to enable a quicker intervention in the event of a fault and optimize the replacement intervals of the grinding belts.

Click here to learn more.



Anomaly Detection: Deep Drilling Process

In a deep drilling process, assessing the stability of the process sufficiently is no easy task. This project examined the possibility of determining the stability using data from accelerometers on the machine together with AI models and mathematical methods. The aim was to optimize the drilling process so as to reduce scrap and material wear.

Click here to learn more.

Prediction: Liquid Distributor Manufacturing Costs

Calculating the manufacturing costs of complex components usually requires a great deal of work. This project used data from a SAP database, AI models and mathematical methods to investigate the possibility of automatically estimating manufacturing costs and other parameters such as the CAD hours or material required.

Click here to learn more.

Embedded, non-obstructive monitoring of voice and speech impediments, taking specific account of privacy

Speech is one of the oldest means of communication. Exclusive to human beings, speech still forms the basis of the digital world we know today, even after around 200,000 years of use. Stuttering is a widely recognized type of speech impediment. It limits the fluency of speech and, in 90 percent of cases, commences before the age of six. At a prevalence of around five percent, the gender split of sufferers shows that the number of male sufferers is double the number of female sufferers. In addition to the outwardly obvious symptoms, stutterers also suffer from insecurity, shame, frustration, aggression, or regression. Self-esteem can also be affected.

Although the medical causes are still unclear even today, behavioural therapy methods have achieved success. One of the therapeutic approaches available is to modify a patient’s speech entirely. In this way, the person concerned initially learns a new way of speaking stutter-free, which is then continually adapted to natural speech. At the start of the treatment, a system of visual biofeedback is used which analyzes the patient’s pronunciation in real time and provides graphical feedback about how the patient uses their voice. A system currently under development attempts to simulate everyday situations using a type of interactive role play. Without leaving the therapeutic environment, the techniques learned are automatically evaluated.

Until now, verification procedures have only taken place during treatment sessions. The “Embedded, non-obstructive monitoring of voice and speech impediments, taking specific account of privacy” research project therefore intends to make it possible to improve ongoing checks, by measuring the use of the newly acquired form of speech in everyday life. The system is designed to operate without an internet connection and to run in the background.

Contact: Prof. Korbinian Riedhammer

Collaborations: Kasseler Stottertherapie (KST)

Fusion of Neural Networks (FuzeNN)

The neural networks currently used for automatic classification use several tens to hundreds of millions of parameters, which places considerable demands on memory and computational resources. The problem intensifies when the networks are evaluated in parallel. Fusion of compressed neural networks is an innovative way of allowing for the concurrent use of parallel networks in embedded systems.

Contact: Prof. Tobias Bocklet

Collaborations: Intel Labs Germany

Identification of security-relevant anomalies in critical infrastructures (ISAKI)

IdSA aims to provide illustrative indications of possible security-relevant processes and situations, with the water supply area of Erlanger Stadtwerke AG (ESTW) as a user, by means of a suitable IT demonstrator. IdSA is targeted towards achieving the earliest possible detection of rare and unexpected anomalies resulting from crime, terrorism, or comparable threats in complex, spatially distributed plants, which either go undetected or are detected (too) late by the control, regulation, and guidance technology of the technical subsystems.
This objective is to be achieved by means of the acquisition, merging, analysis, and classification of both structured and unstructured data and information, as well as their temporal sequences.

Information from public situation reports and open sources regarding changes in the risk situation with a long early warning period is also taken into account. IT security aspects are not at the forefront of the research; however, the solution should be capable of taking account of information from IT security components.

The project aims to significantly improve the state of the art and to lay down the foundation for the subsequent use of the findings, as well as to ensure connectivity in the following areas: science, teaching, economic exploitation.

Contact: Prof. Korbian Riedhammer

Collaborations: ESTW Erlanger Stadtwerke AG

Autonomous mobile measuring and mowing robot for testing open-space PV systems (AutoPV)

In the AutoPV project, the project partners are aiming to develop a system for the autonomous inspection and maintenance of open-space photovoltaic systems. Their goal is to create a functional prototype which, with neither human supervision nor remote control, travels around an open-space photovoltaic system, analyzing the solar modules using a range of measurement methods and cutting back unwanted vegetation.

The prototype is designed as a moving robot, which orients itself by means of satellite navigation, inertial sensors, colour cameras, and a 3D laser scanner. An infra-red camera is used for thermographic inspections to detect faulty modules. Motorized scythes installed on the front allow the area under the solar modules to be maintained while also ensuring that the robot is able to progress unhindered.

 

Contact: Prof. Jörg Arndt

Collaborations: University of Würzburg, Helmholtz Institute Erlangen-Nuremberg, AutarcTech GmbH, Optris GmbH, Evocortex GmbH, Suntec Energiesysteme GmbH

Emotion AI – Is the most compassionate computer the best computer?

The “Emotion AI” project is based on the idea that natural interactions with digital technology can be better adapted to individual people if the technology responds to the emotional state of the user. A technical buzzword in this context is “affective computing”: this aims to make human-computer interactions more empathetic. Computers should not only be able to process explicit human signals, such as touchscreen gestures, clicks, text entries, or speech commands, they should also understand the emotional context of an interaction and react accordingly. The feedback from the computer on the user’s emotional state allows for “Emotion Artificial Intelligence”: emotional intelligence from the computer/machine.

After an intensive planning phase, the project team chose to work more intensively on the case of a “car ride”. The research revolves around the question of how smart technology can support or accompany drivers in different scenarios and situations. The focus is on recognizing and interpreting the driver’s emotional state through the use of AI and optimized measurements. This centers around warning and information sounds and their influence on comfort and safety, including with regard to the user experience of drivers. Because one thing is already clear: the car of the future will respond to the needs of its users and become a personal assistant with tailored suggestions and support for all kinds of situations.

 

Contact: Prof. Yves Ebnöther, Prof. Alexander Hahn, Prof. Alexander von Hoffmann, Prof. Markus Kaiser, Prof. Tilman Zitzmann

Collaborations: tawny.ai, Innovationsagentur HYVE AG

This project is supported by LEONARDO – Center for Creativity and Innovation, a collaboration between Nuremberg Tech, the Academy of Fine Arts, Nuremberg, and the Nuremberg University of Music. LEONARDO is financed by the Federal and State Government “Innovative Hochschule” programme.

EMPAMOS – Game design elements

Since 2016, the Faculty of Computer Science has been conducting the EMPAMOS (Empirical Analysis of Motivating Play Elements, https://empamos.in.th-nuernberg.de/) research project in collaboration with the city of Nuremberg. This project investigates how game design elements can be used in non-gaming contexts to motivate human behaviour in a playful way (“gamification”). For this purpose, the more than 30,000 game manuals held at the Nuremberg German Games Archive are being digitized and examined for motivationally significant game design patterns. Machine learning methods are also used alongside qualitative methods, which has resulted in a catalogue of 104 game design elements and 1,880 described element connections, which is currently based on around 44,000 pieces of empirical evidence.

More specifically, the following three products or services have reached product maturity:

The analogue Game Design Toolbox contains a selection of game element cards compiled for the respective application context, which can be arranged in networks so that a variety of game-based motivational solutions can be generated on the table in team workshops. While the advantage of the analogue box is the intuitive haptic interaction with the material during the creative process, the digital toolbox allows for quick access to the more than 1,880 element connections described.

The analogue and digital Game Design Toolboxes were applied in the fields of corporate and IT consulting and cultural administration (companies, social work, cultural institutions, games industry), corporate and IT consulting.

 

Contact: Prof. Laila Hofmann, Prof. Robert Lehmann, Prof. Thomas Voit, Prof. Rainer Kotzian

Collaborations: Nuremberg Games Archive, Research and Innovation Laboratory for Digital Teaching (FIDL)

This project is supported by LEONARDO – Center for Creativity and Innovation, a collaboration between Nuremberg Tech, the Academy of Fine Arts, Nuremberg, and the Nuremberg University of Music. LEONARDO is financed by the Federal and State Government “Innovative Hochschule” programme.

Smart form completion assistant for supporting homeless people (INA)

INA is a virtual, discreet, and smart assistant for social workers. First, it identifies possible benefits in a dialogue with the people affected, then it helps fill in the necessary forms correctly and prepare the supporting documents needed.


Applying for social benefits often presents citizens with unexpected difficulties, whether due to ignorance, shame, or language and understanding barriers. It is often necessary to translate the bureaucratic language into simple German and vice versa. Social workers play an important role in this regard. However, filling out forms together is time-consuming and not one of the core elements of social work. Moreover, social workers cannot reach all citizens in need. The intention is for INA to act as a virtual assistant, taking care of simple cases automatically as far as is possible in order to relieve professionals, to free up space for counseling and relationship work, and to break down barriers.

INA is to be implemented as a multimodal virtual assistant, initially with chat and speech as its input and output. AI technologies play a central role in this regard: for example, forms are to be prepared with the help of ontology and text recognition for a statistical dialogue manager, which in turn is to be trained on real counseling dialogues for supporting homeless people. Speech recognition and language processing (NLP), coupled with speech synthesis, allow the users to be understood, and speech synthesis makes a successful dialogue possible through the use of statements and follow-up questions. INA can be used upstream of the actual counseling sessions, thereby reducing the burden on social workers.

The focus is on social services in the context of homelessness, since rough sleepers and homeless people tend to be very suspicious and fearful of bureaucratic administrative actions. However, the idea that a virtual assistant can learn any forms together with the corresponding dialogue guidance means that its application is transferable to almost any issues encountered in interacting with the authorities. Due to its low acceptance threshold and constant availability, areas such as addiction, unemployment, and hidden poverty can benefit greatly from discrete assistance. In the long term, our approach could be established as a complement to “traditional” counseling services.

The applicants are very well networked through numerous research projects on homelessness in the region and can draw on the data they have gathered on the needs and realities of homeless people (SIWo: smart inclusion for homeless people). The dialogue guidance for identifying and applying for benefits is learned on the basis of recordings of counseling interviews in various facilities that assist the homeless, which arise as a result of real support situations with affected persons. The project is supported by the Stadtmission Nürnberg, Don Bosco Jugendwerk Nürnberg, and Caritas Regensburg, which ensure access to counseling for homeless people as well as the diversity of the target group.

 

Further information:

https://www.civic-innovation.de/das-foerdern-wir/praemierte-ideen-1-runde

https://www.civic-innovation.de/service-und-beratung/aktuelles-und-termine/standard-titel/ideen-fuer-gemeinwohlorientierte-ki-anwendungen-erste-preisverleihung-der-civic-innovation-platform

Jointly towards AI – Nuremberg Institute of Technology (th-nuernberg.de)

 

Contact: Prof. Korbinian Riedhammer, Prof. Frank Sowa

Collaborations: Caritas Regensburg, Don Bosco Jugendwerk Nürnberg

Interactive Artificial Intelligence (IAI)

Generation Z and the subsequent generation Alpha are growing up with a mobile phone or tablet in their hands. While the previous generation used to talk to friends or write to Dr Sommer at the youth magazine, Bravo, about embarrassing topics, nowadays young people simply have to reach for their mobile phone – this opens up a multitude of possibilities for obtaining information: a phone call or a message to a friend, posting a message in a forum where they can exchange ideas with “like-minded people” or accessing secure mail counseling from services such as www.sextra.de by profamilia, where responses are provided by a trained counselor.

But what about young people who would rather remain anonymous when it comes to embarrassing topics? This issue is addressed by IAI (Interactive Artificial Intelligence). In this project, a prototype chatbot is to be developed that interacts with its users on these difficult topics in an easily accessible manner. More specifically, it adapts to the users by analyzing the course of the conversation in real time and adapting its response behaviour accordingly. Its answers are therefore increasingly better adapted to the needs of the person seeking advice. The chatbot is in no way meant as a substitute for a professional (online) counseling service, but to act as a supplement and support for both counselors and those seeking advice.

The consortium, which consists of representatives from the Faculties of Social Sciences, Business Administration, and Electrical Engineering, will develop various technologies in the field of machine learning for this chatbot during the course of the project, and train it to develop knowledge of potential users. This will ultimately lead to the development of a prototype that can then undergo a test run with a selected group and interact with them.

 

Contact: Prof. Robert Lehmann, Prof. Thomas Bahlinger, Prof. Oliver Hofmann

Collaborations:  Pro Familia with online counseling: Sextra

This project is supported by LEONARDO – Center for Creativity and Innovation, a collaboration between Nuremberg Tech, the Academy of Fine Arts, Nuremberg, and the Nuremberg University of Music.
LEONARDO is financed by the Federal and State Government “Innovative Hochschule” programme.

Can algorithms calculate morality in situations of conflict? (KAIMo)

The project attempts to find answers to the following research questions:


Philosophy: How can the normative aspects of child welfare risk be conceptualized (e.g. child welfare, self-determination, fair procedures, avoidance of discrimination)? How can the resulting normative criteria be translated into algorithms, into rational decision-making processes, and beyond that into digital social work practice? What understanding of normativity, rationality, and values is suitable for this and what are the consequences of the practical implementation requirements for philosophy and its understanding of normativity in the digital age?


Social Work: How can work processes and decisions in social work be supported digitally? How can the normative aspects of decision-making be formulated appropriately? How can appropriate assessments be made of child welfare risk and the evaluation of risk potential on this basis? How can ethically grounded digital support practices help with this?


Computer science: How can normative criteria be quantified? How can normative dimensions be translated into algorithms? What specific peculiarities of ethical conflict cases do we need to pay particular attention to? What are the consequences of the findings and requirements in terms of computer science for the other two disciplines? How can algorithms contribute to decision-making through scoring?

 

Contact: Prof. Robert Lehmann

Collaborations: Munich School of Philosophy, Würzburg Schweinfurt University of Applied Sciences

KI-Power – System for the flexible testing of model-based predictive and artificially intelligent control methods in electrical drive engineering

An innovative, flexible, modular, and high-performance heterogeneous platform for novel control concepts in the field of power electronic systems will be investigated within the scope of the KI-Power research project. The focus here is on real-time capability, a broad range of applications, and high computing power. Despite promising research results, the industrial implementation of cutting-edge approaches to the control of power electronic systems, such as the use of methods from the field of artificial intelligence (AI) or model predictive control (MPC), is still pending. Reinforcement learning, an AI method, and MPC have numerous theoretical similarities and connections and are sometimes used in combination. Similar challenges arise when it comes to the industrial implementation of these methods in the field of power electronic systems. Both methods require high computational power to meet the real-time requirements of these applications. Powerful computational platforms coupled with the ability to use these to develop products through rapid prototyping are therefore key in establishing AI in the industry.

State-of-the-art System-on-a-Chip (SoC) solutions, which combine multiple processors and Field-Programmable Gate Arrays (FPGA) in a single chip, provide the necessary computational power for both AI and MPC and will be referred to as SoC-FPGA below. The high degree of complexity of SoC-FPGA and the lack of experimentally validated implementations of MPC and AI processes currently present a stumbling block for their use, particularly by SMEs. The platform researched within the framework of KI-Power effectively removes this obstacle to innovation and paves the way for the development of a new generation of innovative systems in the field of power electronics and electrical drive engineering. KI-Power offers a hitherto non-existent, holistic view of trustworthy, microelectronic hardware and software components as part of an innovative platform. In the hardware domain, this is characterized by minimized latency, minimized jitter, optimized signal integrity, and enormous computing power. This computing power results from the heterogeneous computation of algorithms with an SoC-FPGA processor family that is scalable in terms of both cost and computing power. As regards software, a modular framework is being explored for the platform, which will make use of a transparent workflow for software and hardware development with the use of open source tools (including PYNQ, Python) and automated code generation.

Since power electronic systems already play a pioneering role in Germany and the implementation and experimental testing of smart control processes for systems of this type present a challenge, KI-Power serves as an “enabler” for these innovations. The exploitation concept of KI-Power is based on a two-step approach that follows the “freemium” business model. As a first step, the outcomes of this project (hardware, software, documentation) will be made available under an open source licence, which should help the platform to gain attention and establish itself in the German research and industrial landscape. In the second step, a spin-off from the university network will be established in partnership with the participating companies, which will take over the maintenance, further development, and distribution of the platform. By mapping a holistic value chain in Germany, the yet-to-be-founded company and the participating companies should benefit from Germany’s position as a technology pioneer.

Contact: Prof. Armin Dietz

Collaborations: Munich University of Applied Science, Trenz Electronic GmbH, Kübrich Ingenieursgesellschaft mbH & Co KG, Afag GmbH

Mixed-reality mining - HoloMine

To achieve acceptance in society and profitability, the mining operations of the future will need to be optimized in terms of their impact on society and the environment, safe and healthy working conditions, and mining processes.

Mine digitalization (Mining 4.0) is identified as a key technology for success, while Mixed Reality (MR) is the most promising way of enabling users to leverage all benefits to improve operational efficiencies, through the integration of real world and virtual world in one 3D/4D environment.

Using existing MR hardware, the development of specific solutions for selected underground mining issues, including an overall enterprise data infrastructure, is planned within HoloMine.

The pilot solutions will provide clear evidence that mixed reality solutions are perfect for enabling rapid shaft and roadway inspection, machine maintenance, 3D model visualization, and quick decision-making. The mine will become a completely new canvas for everyone to understand and explore, to learn, communicate, and interact with.

 

Contact: Prof. Stefan May

Collaborations: DLR, DMT, LTU - Business AB, EIT - European Institute of Innovation & Technology, Microsoft HoloLens, University of Leoben, robotic eyes, Sandvik, Freiberg University of Mining and Technology, Graz University of Technology

New paths in robot programming with human-robot collaboration (HRC) and motion capture (MoCap) - MRK&MoCap4Robots

Modern industrial robots are now able to perform their working steps in direct cooperation with humans. This means that robots can be used very flexibly in production – even in mixed plants with humans and machines. However, very considerable expense is incurred for programming, which requires time and expert knowledge. In order to reduce the time expended and to simplify the programming of robotic systems, a method toolbox is to be developed at Nuremberg Institute of Technology over the course of a four-year research project for use in practice. This will allow industrial robots to be programmed easily and in a short time with MoCap, AR, and VR and, depending on the programs, to perform multiple tasks.

 

Further information: https://mediasharing.service.th-nuernberg.de/video/Intuitive-AR-MoCap-Robo-Teaching/df70e028bf132f7f9ad97f603d0040bc

Contact: Prof. Michael Koch

Spirio Sessions

Nuremberg Tech and the Nuremberg University of Music are working together on the “Spirio Sessions” interdisciplinary research project concerning artificial intelligence, or more specifically, artificial creativity. The Interdisciplinary Music Research department at the Nuremberg University of Music – which recently saw success in the AI competition run by the Bavarian State Ministry of Science and the Arts with a professorship for “Artificial Creativity and Musical Interaction” – is working together with machine learning scientists at Nuremberg Tech’s Faculty of Computer Science.

The focus of this joint research is on developing “artificial creativity”: the interaction between human musicians with innovative musical instruments and artificial intelligence. At the “heart” of this joint research project is the self-playing Spirio R grand piano – an analogue-digital hybrid instrument from Steinway & Sons. The researchers hope to use the instrument not only to gain data-based insights into musical interpretation but also to investigate in particular the possibilities of an interface with artificial intelligence.

It may soon be possible to answer questions such as: Can the grand piano improvise in a jazz quartet? Even adding its own personal touch? And what insights will this give us into the supposedly exclusively human creative process? The project team hopes that the results will also provide insights that can be transferred to other areas and application scenarios – such as medicine.

 

Further information: https://leonardo-zentrum.de/projekt-spirio-sessions/

Contact: Prof. Korbinian Riedhammer

Collaborations: Nuremberg University of Music

 

This project is supported by LEONARDO – Center for Creativity and Innovation, a collaboration between Nuremberg Tech, the Academy of Fine Arts, Nuremberg, and the Nuremberg University of Music.
LEONARDO is financed by the Federal and State Government “Innovative Hochschule” programme.

Underground Robotic System for Monitoring, Evaluation, and Detection Applications (KAVA UNDROMEDA)

The number of underground mines is increasing worldwide. The hazardous environments in mines result in a high risk of injuries and fatalities. The approximately 657 underground mines in the USA alone have an average of 20 fatalities and 1,972 serious injuries per year, which are mostly caused by ceilings or walls collapsing. Therefore, the demand for exploration of unknown areas for operational and safety purposes in underground mines, but also in sewage tunnels and other underground areas, is very high today and will increase in the future. The UNDROMEDA project aims to develop a robotic underground measurement system for autonomous 3D mapping and monitoring. The system is based on a mobile, wheel-driven platform which additionally carries a flying drone to approach areas in underground mines and other underground environments that are particularly unknown, difficult to access, or hazardous. The autonomous platform and drone will significantly reduce the risk to underground personnel by replacing manual measurements. Automation will reduce the time and expense needed for mapping and monitoring while advanced sensors and their integration will dramatically enhance the information density and quality.

 

Further information: Datasheet.pdf

Contact: Prof. Stefan May

Collaborations: Boliden Mineral, Caterpillar Global Mining Europe GmbH, French Alternative Energies and Atomic Energy Commission, CEA, Deutsches Zentrum für Luft- und Raumafahrt e.V., DMT GmbH & Co. KG, Indurad GmbH, Inkonova AB, LKAB Minerals AB, University of Leoben, RWTH Aachen University, RISE Research Institutes of Sweden AB, Freiberg University of Mining and Technology.

Unkafka – Making official letters simple and understandable

Unkafka is an AI-based assistant that automatically recognizes complicated formulations in official letters, marks them visually, and generates suggestions for simplifications. It supports clerks and lawyers in the preparation of official notices and ordinances. The suggestions are generated by a model that is trained specifically for this task using available training data. Unkafka is to be made available both as open source software and as a directly usable service.

The idea aims to bring about sustainable improvement in the social acceptance of official and governmental decisions. Many citizens feel abandoned when it comes to decisions and regulations and overwhelmed by the speed and extent of changes. Unkafka helps people to better understand official correspondence. The partners want to bring their extensive experience in the governmental sphere to bear in the form of a vision for improving and optimizing dealings with citizens with a view to bringing about sustainable change. According to their assessment, added value for society and the authorities alike can be created by simplifying the preparation of easily understandable notices. In addition, there is a great deal of self-interest in practically implementing or supporting useful projects and, where these succeed, benefiting financially from their implementation. The results are also intended to be used in research and teaching.

 

Further information:

https://www.civic-innovation.de/das-foerdern-wir/praemierte-ideen-1-runde

https://www.civic-innovation.de/service-und-beratung/aktuelles-und-termine/standard-titel/ideen-fuer-gemeinwohlorientierte-ki-anwendungen-erste-preisverleihung-der-civic-innovation-platform

Jointly towards AI – Nuremberg Institute of Technology (th-nuernberg.de)

 

Contact: Prof. Jens Albrecht, Markus Stadi (Federal Employment Agency)

Collaborations: Federal Employment Agency

Validation of artificially intelligent control methods for electric drive system technology – KIRA

The KIRA project aims to reduce the vibration load on electric drive systems
and to simplify drive system controller settings. The idea is to achieve this through the use of an artificially
intelligent control method (AI), which makes use of reinforcement learning
methods. As part of the project, the functional principle and the advantages of AI control will
be validated. For this purpose, a proof-of-concept of the method will be provided and its advantages quantified against
the state of the art. For the AI control, an IP core is to be developed that can be integrated into
standard frequency converters, thereby allowing for an accelerated market introduction of the process. It is envisaged that this use will bring about the spin-off of a company for the commercialization of the
AI control and the initiation of further research projects.

 

Contact: Prof. Armin Dietz

Fully automated shunting locomotive for Deutsche Bahn (DB Cargo)

In the VAL (fully automated shunting locomotive) project, a retrofit module for a shunting locomotive is being developed and brought to application maturity. The module allows for fully automated shunting without the presence of a locomotive shunting driver. Individual components from other branches of industry are modified and combined to form an innovative, novel system. In addition, a new digital interface will be created between the infrastructure and the locomotive, which will provide an adequate replacement for the current verbal communication. A new computer component will also be integrated into the signal box, which will be used to transmit driving commands to the locomotive. The result is a product that is ready for application, that fulfils all the necessary safety-critical requirements, and that enables a series release with market launch.

For the validation and evaluation of the functions that are to be automated in order to drive the fully automated locomotive, specific tests, which include relevant driving manoeuvres under environmental influences, are necessary in order to compare the reaction of the overall system to the respective situation against what is expected of it. As a result, Nuremberg Tech is focusing on developing methods and criteria for testing functional requirements from the point of view of perception, localization, classification, and the resulting driving and braking decision.

For the performance of the tests, two test environments will be created in parallel, within which a systematic and consistent test procedure will be developed and applied. A laboratory environment and a characteristic real-world environment in the field will be created as test environments, in which the system’s reactions to defined disturbance and input variables will be evaluated on the basis of defined criteria. Traditional software verification methods such as regression testing are used, together with verification methods for software with high AI (artificial intelligence) content, such as Adversarial Examples/Attacks or Explainable Artificial Intelligence techniques.

The creation of test environments and scientific tests that are independent of the development of the system leads to increased system security, since the requirements cannot be described precisely due to the large number of possible situations, but still have to be evaluated according to well-defined criteria.

The project is funded by the Federal Ministry of Transport and Digital Infrastructure.

 

Contact: Prof. Cichon, Prof. Tavakoli Kolagari, Prof. May

Collaborations: DB Cargo AG

Project duration: December 2020 to December 2024

Completed projects

Computer-aided analysis of social science texts using machine learning techniques (CASoTex)

The aim of this project was to investigate the possible applications of computational linguistic methods and machine learning techniques for the automated analysis of social science texts. The data basis was made up of contributions from moderated, social consultation forums in digital format. On the basis of specific social science-related questions regarding interdependencies in e-counseling, the project aimed to investigate the extent to which computational linguistics and machine learning techniques can support or complement qualitative analyses, where the limits of these techniques lie, and how to proceed when using them. The project generated initial concrete findings, the research network expanded, practical research projects were applied for, and further considerations for larger research projects emerged. The results so far are promising. With a sufficiently large amount of well-prepared training data, supervised learning methods can provide interpretable and usable findings.

 

Further information:

Lehmann, Robert; Albrecht, Jens & Zauter, Sigrid (2020): CaSoTex, Computerunterstützte Analyse Sozialwissenschaftlicher Texte. E-Beratung, 2020 (1). doi: 10.34646/thn/ohmdok-619 . https://opus4.kobv.de/opus4-ohm/files/619/CaSoTex-Abschlussbericht2.pdf (pdf 464 KB)

Grandeit, Philipp; Haberkern, Carolyn; Lang, Maximiliane; Albrecht, Jens & Lehmann, Robert (2020): Using BERT for Qualitative Content Analysis in Psycho-Social Online Counseling. 4TH WORKSHOP ON NLP AND CSS at the EMNLP 2020 The 2020 Conference on Empirical Methods in Natural Language Processing. https://www.aclweb.org/anthology/2020.nlpcss-1.2/

Lehmann, Robert; Albrecht, Jens & Zauter, Sigrid (expected 2021): Die Computerunterstützte Analyse Sozialwissenschaftlicher Texte – Ergebnisse des Forschungsprojekts “Casotex” in der psychosozialen Onlineberatung. In: Carolin Freier; Joachim König; Arne Manzeschke & Barbara Städtler-Mach (publisher): Gegenwart und Zukunft sozialer Dienstleistungen. Wiesbaden: Springer.

 

A collaborative project by the Institute for E-Counseling, Faculty of Social Sciences, Prof. Robert Lehmann, and the Faculty of Computer Science, Prof. Jens Albrecht, funded by Nuremburg Tech’s preliminary research fund for 2019.

Digital assistance in psychosocial counseling (DiA)

The aim of the project was to develop digital, natural-language assistance in the form of the “DiA” chatbot in collaboration with the Institute for E-Counseling at Nuremberg Tech. The plan was to use educational counseling in the context of the certificate course in E-Counseling as a specific test field, together with psychosocial e-counseling. The aim was to use an operational prototype to investigate the extent to which modern chatbot technology, backed up by artificial intelligence methods, can be effectively used for personal counseling in the aforementioned environment. At present, artificial intelligence is rarely used for counseling in the social sector, although there is a great and urgent need for communication that is accessible at all times, especially in this area.

The project ran from September 2019 to February 2021 and was funded by the STÄDLER Foundation.

 

Further information: DiA in Ohmjournal 2020/02 (pdf) (pages 82-85)

Contact: Prof. Robert Lehmann

Digital empathy in user experience research – development of an integrated affective computing measurement method for research into digital attention and emotion

The aim of this project is to complete a feasibility study and preliminary work for an integrated affective computing measurement method for research into digital attention and emotion. This should consist of a combination of eye tracking and facial coding. This solution is innovative for two reasons: First, it takes account of the latest neuroscientific findings on the interaction between attention and emotion when using (digital) products. Furthermore, it transfers affective computing methods into the field of application for user experience [UX] research, where there is a very large gap in application-oriented research for industries such as digital media, communication, and mobility.

Contact: Prof. Alexander Hahn