Research Topics

We focus on three research vectors (RV1, RV2, and RV3) as described below.


Safety is a critical research aspect for automated driving. To ensure and quantify safety, it is important to develop new solutions at various system-levels (from algorithms to platforms, from concepts to metrics). In this regard, research questions include:

  • Technical realization of live monitoring from an autonomous perspective to ensure safety of vehicles under physical constraints, ensuring naturalistic driving. This builds upon the concept of RSS (Responsibility-Sensitive Safety).
  • Safety in automated driving systems and algorithms, decision making for driving policy, reinforcement learning.
  • Aspects of technical implementations: Economic scalability, fast computation, platforms, and new algorithms.
  • Aspects of new sensor fusion approaches considering new novel safety assurance aspects; statistical models and (formal)verification method.
  • Regulatory compliance aspects.


Validation aims to prove that the techniques for AV meet the requirements of the users. In the context of systems using machine learning this task is harder as, for example, classical software verification and validation tests cannot directly be applied. Therefore, new methodologies need to be developed that also consider the data used for training and testing. Techniques like simulation are promising, but their effectiveness still needs to be demonstrated. Research questions include:

  • Techniques for the creation of realistic training and validation data; modeling of human appearance and behavior.
  • New approaches to tackle the complexity of systems (and sub-systems) vs. completeness or coverage in validation.
  • Benchmarksusing real data sets for a consolidated validation strategy.
  • Validationthrough physical-based simulation techniques and development of accurate sensor and scene models for Lidar, RADAR combined with vision-based sensors (cameras) in one simulation system.
  • Approaches to achieve and ensure more robustness of AV algorithms.


For a comprehensive safety solution that allows optimization for aspects of safety, comfort and global traffic flow, it is important to not only consider a single vehicle and its in-vehicle perspective. Instead, the interaction between diverse traffic participants and infrastructure has to be included in the development and research. This includes the consideration of connected and unconnected AVs, Non-AVs, humans, and communication connections between those entities. Anend-to-end safe sensing approach requires consideration of failures in communication, computation, and sensors as well as different weaknesses in different sensing modes, perspectives, and positions. Additionally, safe automated driving approaches shall consider the level of trust of different entities.

  • As traffic involves usually more than one system at a time, this RV extends to multiple connected agents and explores safe sensing strategies for ego and infrastructure perspectives.
  • Extension of existing or evolving sensing strategies to include novel fusion techniques of combinations of different sensing strategies (lidar, radar, vision-based) of different sensing positions and perspectives (ego and infrastructure) considering advantages of different modes and perspectives.
  • Dependable communications, architecture, protocols for dynamic scenario management based on advanced diagnostics (e.g., sensor, communication and compute failures; sensing deficiencies like weather, blind spots because of very dense traffic, single measurement failures or complete component malfunction).
  • Establishing the level of trust of participating entities (AV, non-AV, and infrastructure) and system scalability and effectiveness.
  • Novel sensing, management, and connection approach for new connected transport approaches (including novel multi-modal transport approaches and novel infrastructure-changing and extending approaches).