Living lab for the transfer of digital health applications and AI into healthcare
Applied Artificial Intelligence
Artificial Intelligence (AI) as a key technology enables and enhances many research projects at the FZI – especially in the application fields of Mobility, Transportation and Logistics, but also in Production, Supply and Disposal, as well as Healthcare.
AI plays a strong role in the FZI’s research on human-robot collaboration, functional safety – especially for robotics and mobility – or sensor and image processing solutions across all subjects.
A particular concern of the FZI is to evaluate the latest AI methods and procedures from research, adapting and transferring them into practical applications at partner companies.
The FZI is particularly adept in the area of Applied Artificial Intelligence concerning criteria of transparency, coherence, and security.
- Research and application competencies regarding the interplay of artificial and human intelligence, human-AI interaction and AI adaptation
- Occupation with knowledge graphs as well as with Explainable AI systems (transparency and coherence)
- Know-how on methods of automatic image understanding and pattern recognition or their interpretation by AI
- Use of AI for the prediction of states, time series, or motion sequences
- Knowledge of dedicated hardware for AI (edge computing, neuromorphic hardware, graphics processing units)
- Expertise in AI engineering for efficient integration and validation in concrete applications or strength in the engineering of AI systems
- Expertise in the verification of AI methods, AI model maintenance and safety & security
- Know-how on Few-Shot Learning, Transfer Learning (i.e., Zero-Shot Transfer as well as Human-in-the-Loop Methods), Curriculum Learning, Reinforcement Learning
- Know-how on data augmentation, data synthesis, data-driven engineering and applied data analytics
Within the Applied Artificial Intelligence research focus, the FZI will spotlight the continuous expansion of the topics of human-in-the-loop processes, embedded neuromorphic computing and self-supervised learning. Enhancing the expertise in the area of evaluating and generating texts in natural language – namely Natural Language Processing (NLP) – is an especially high priority. Topics such as Reinforcement Learning and knowledge representation are included increasingly.
The goal is to develop representations of knowledge, such as language, which can be used by humans as well as for AI methods. This is necessary, for example, in the automotive sector, in media and communications (i.e. for the analysis and recognition of disinformation campaigns), or for Industry 4.0. Furthermore, end-to-end learning methods help improve the compliant control of robot arms. Moreover, the reactivity of robot systems can be increased efficiently through the skilful use of pre-learned movements.
What distinguishes the FZI’s work with Artificial Intelligence is its closeness to the application and its strength in engineering AI systems. Approaches are always based on systematic thinking, from the AI function to the realisation. For the FZI, applying AI in embedded systems with restrictions in terms of resources, energy, or interfaces is a focus. It is essential to include the constraints of the hardware boundary conditions in a consistent design during the early development stages. The FZI benefits from its well-equipped laboratory environment with real application scenarios for practical relevance.
Digital Hub Karlsruhe, AI Regio, FZI AI Day