AI-Methods and data management

Artificial intelligence (AI) methods are increasingly applied in mobility systems, i.e., for functional components in highly automated cars, intelligent infrastructure, mobility apps, or in backend systems. We address the improvement,  applicability and validation of new machine learning methods for interpretation, perception and planning. An additional aspect is the system integration in terms of hardware and software components.   

Data plays an increasingly important role in function development and optimization as well as in their validation. Especially AI and components based on machine learning need huge amounts of linked data on different levels of abstraction and from different sources. Based on use cases analysis, we address the data acquisition and aggregation, data analysis and consistency, data privacy, concepts for architectures, and metadata schemes. In addition to using real data, we are also working on virtual reality and simulations to support research and development of mobility systems. Emerging machine learning methods that require less data are also being investigated and further developed.

Safety & Security

The area address research towards functional safety of automated mobility systems but also security of subsystems and their functional components.  Another aspect concerns the detection of attacks (intrusion detection or adversarial attacks) and protection of distributed and networked sub systems. To prevent and repel these attacks, we have developed realistic attacker models that are based on historic, real attacks on mobility services.

Models for safety of today’s mobility system have to allow for dynamic reconfiguration with high variability, which on the other hand means that there is a higher amount of uncertainty at development time. Furthermore, these systems are long-living and have to be tested and verified for safety, security and dependability. To estimate the quality of system verification process, we have developed a model for that can be used to identify the challenges that this type of uncertainty poses.

Quality in Systems of Systems Engineering

In mobility systems, there is a wide spectrum of sub-systems which have to interact. Since many of these systems are developed independently of other parts, it is paramount to apply quality management methods that can describe multiple heterogeneous systems. Modern development processes use model-driven methods to describe systems at a high level of abstraction. In this research area, we make use of these models to describe inter-model consistency with specialized domain-specific languages. Consistency preservation mechanisms help system developers and maintainers to semi-automatically repair models after changes have happened, which is a considerable advantage over state-of-the-art processes, where these consistency checks are performed manually. This way, shorter development lifecycles can be reached and updates of critical infrastructure can be performed with shorter reaction times.

Automated driving and connected infrastructure

Highly automated and autonomous vehicles need to be able to sense the world, predict the behaviors of other traffic participants, act cooperatively, and plan safe trajectories in real-time. In this research area, we develop methods for everyday driving conditions as well as corner cases and integrate them in fail-safe architectures. In addition, we are developing new methods for X-in-the loop validation, for example based on virtual reality. Research results are evaluated on test vehicles in real road traffic, e.g., autonomous cars for individual mobility and shuttles for on demand mobility. Intelligent infrastructure will increasingly support mobility systems. We investigate new sensor concepts for the infrastructure, advanced communication protocols and edge computing and distributed systems for the control of autonomous vehicles.  

Development methods for interdisciplinary collaboration

Research on digitalization for more sustainable mobility covers numerous disciplines, including engineering, natural sciences, social sciences or design. Interdisciplinary collaboration that addresses the topic from different perspectives is the key to developing solutions for society as a whole. In this research area, we address practices and challenges of collaborative, digital cooperation for improved interactions, decision-making, and productivity.