Pioneering research at the interface of software engineering sciences and AI
Despite major technological advances and the growing adoption of AI-powered autonomous systems, trust in their safety, reliability, and transparency remains a key concern. The development of a new generation of robust AI technologies is therefore essential for a large number of complex applications - systems that can make rapid, reliable and comprehensible decisions in uncertain and unpredictable environments. These must not only process large amounts of data efficiently, but also extract valuable and accurate insights from limited data sets without jeopardising confidentiality and privacy.
AI engineering is gaining importance in an increasingly diverse range of industries as companies face the challenge of transferring AI technologies into value-adding applications in a targeted and sustainable manner. This demanding development process is particularly complex in terms of security, as conventional software engineering methods are often not directly transferable to AI models or AI-based systems. A crucial prerequisite for success is the precise and efficient use of data - whether for training, optimising or validating AI models. Data not only forms the basis for the modelling and continuous further development of AI systems, but also ensures their quality, efficiency and trustworthiness.
The AI Engineering research focus at fortiss concentrates on the development of robust and trustworthy AI technologies that can make fast and reliable decisions in uncertain and unpredictable environments. The focus here is on increasing the trustworthiness and explainability of AI systems in order to make them resilient to incorrect inputs and targeted attacks. Another aim is to process large amounts of data efficiently and at the same time gain valuable insights from small amounts of data without jeopardising confidentiality and privacy.
fortiss actively contributes to the further development of AI engineering by researching innovative approaches such as generative models for data synthesis, human-centred design for improved usability and methods for trust calibration in AI-supported decisions. In addition, fortiss is developing efficient learning methods for resource-limited environments, low-energy hardware solutions and edge and mobile AI concepts that enable decentralised processing and low latency times. In safety-critical areas, testing, verification and monitoring techniques are used to make AI models reliable and traceable.
The research focus on AI engineering pursues a holistic approach that ranges from data analysis and processing, generation and modelling to validation, optimisation and integration into industrial systems. This comprehensive approach ensures that AI solutions are not only powerful, but also reliable, efficient and practical to use. This includes simulation and evaluation environments that integrate generative AI, as well as hardware-adapted software models for edge AI and continuous learning methods. The areas of expertise Neuromorphic Computing, Human-Centered Engineering, Machine Learning and Software Dependability contribute significantly to the development of innovative and practice-oriented solutions.
Development of solutions involving data and knowledge
The Machine Learning (ML) area of expertise focuses on the development and application of advanced, data-driven models that can solve complex problems in challenging areas such as autonomous driving, medical diagnostics and predictive maintenance. A particular focus is on reinforcement learning and representation learning techniques with the aim of optimizing the adaptability of models and increasing their robustness to changing conditions.
The expertise of fortiss covers critical application areas such as image and speech processing, autonomous navigation and recommendation systems. Through the “One Stop Shop” Machine Learning Lab (One-ML), fortiss offers a customized range of services - from in-depth consulting and technical training to publicly accessible lectures. This promotes the transfer of knowledge to industry, education and society and enables the efficient implementation of ML technologies in practice.
The third generation of artificial neural networks
In the field of neuromorphic computing, our activities concentrate on the development of energy-efficient, low-latency neural networks, which are used in robotics and industrial applications in particular. The focus is on researching pulsed neural networks that enable particularly efficient information processing thanks to neuromorphic hardware.
In the field of AI engineering, fortiss develops software methods and algorithms that improve the learning ability and intelligence of technical systems in manufacturing, the automotive sector and robotics. Neuromorphically controlled robotics, which efficiently controls mobile and industrial robots with low energy consumption and low latency times, is particularly relevant.
Another research topic is the development of neuromorphic vision systems for robotics and human-machine interaction. These technologies help to optimize the integration and performance of AI systems in safety-critical and resource-constrained environments.
Understand and explain decisions of intelligent systems from the user's point of view
The Human-centered Engineering competence field at fortiss is dedicated to the development of AI systems that not only meet technical requirements, but also focus on the needs and expectations of users. In the field of AI engineering, there is a particular focus on human-machine interaction. fortiss develops data-based, intelligent user interfaces that enable the natural and intuitive use of AI systems. Particular emphasis is placed on user-friendly, efficient and secure interaction.
Another focus is the transparency of decision-making processes in order to strengthen trust in AI-supported systems. This is achieved by developing models and interfaces that present AI decisions in an understandable and comprehensible way. The aim is to create practical solutions that promote the acceptance and integration of AI technologies in various sectors such as aviation, the automotive industry, healthcare and production.
Rigorous validation and verification for dependable and safe software systems
The Software Dependability (SD) competence field at fortiss focuses on the development of innovative methods, algorithms and tools to ensure the reliability and safety of cyber-physical systems (CPS), in particular adaptive, autonomous systems based on artificial intelligence (AI). These systems learn from data and experience and adapt their behavior to dynamic environments. One of the biggest challenges is to reliably verify and certify these adaptive systems in safety-critical areas such as the automotive and aerospace industries.
fortiss develops advanced verification methods, in particular for neural networks and AI-based systems, to enable continuous validation and testing of software in compliance with relevant standards. Especially in the area of safe and efficient software verification, formal methods such as model checking and static analysis are combined with other test procedures. This ensures the reliability and safety of autonomous CPS and specifically promotes the use of AI in safety-critical applications.
By continuously analyzing high-frequency sensor data, fortiss enables the early detection of anomalies in satellite constellations. Advanced pattern recognition and deviation analysis algorithms identify potential failure sources, helping to prevent system malfunctions and ensure operational stability.
Machine learning supports companies in making production processes more efficient and predictive. fortiss develops adaptive models that forecast maintenance needs, detect quality deviations in real time, and optimize workflows. This reduces downtime and costs while increasing overall productivity.
Intelligent control algorithms enable energy grids to dynamically adapt to actual demand. fortiss explores methods for optimal load distribution, facilitating the seamless integration of renewable energy sources while maintaining grid stability. This approach supports a reliable and sustainable energy supply.
fortiss develops neuromorphic computing architectures for ultra-fast LIDAR data processing, enabling precise environmental perception and reliable obstacle detection in autonomous aerial systems. This high-performance real-time analysis enhances navigation capabilities and increases safety in complex flight environments.
Mit innovativen KI-Methoden ermöglicht fortiss die Echtzeitanalyse biometrischer Daten auf tragbaren, energieeffizienten MedTech-Geräten. Durch neuromorphe Verarbeitung werden Vitalparameter kontinuierlich erfasst und ausgewertet, um frühzeitig gesundheitliche Auffälligkeiten zu erkennen. Diese Technologie unterstützt sowohl Patient*innen als auch medizinisches Fachpersonal bei präventiven und diagnostischen Maßnahmen.
How can secure software systems be developed with AI? This whitepaper explores technical design and engineering principles for secure AI systems. Using an emergency braking system as an example, it highlights key challenges in specification, safety, design, and maintenance.
This whitepaper introduces Artificial Intelligence (AI), explores its application areas, and highlights its impact on economic growth. It analyzes success strategies, Germany's position, and provides recommendations for the economy of Bavaria.
How are secure software systems with AI developed? This whitepaper covers technical design and engineering principles. Using an emergency braking system as an example, it explains key challenges related to specification, safety, design, and maintenance.
Download pdf
This whitepaper discusses knowledge-enhanced machine learning (ML) and solutions to the limitations of current deep learning models, such as lack of adaptability and task specifications. It explores invariance and learning paradigms that enable autonomous knowledge acquisition.
Autonomous cognitive systems promise higher performance in complex situations. The VDE-AR-E 2842-61 standard provides a risk-based approach to ensure the trustworthiness of AI systems during development and operation.
Download pdf
Artificial Intelligence is a central topic in computer science. This whitepaper explores AI application areas, their impact on economic growth, and strategies of successful nations, with a focus on Bavaria.
![[Translate to English:] Graphik AI Engineering](/fileadmin/user_upload/04_Forschung/fortiss-cluster-AI-engineering.png)
