Learning from Humans – Building for Humans

Project type: Funded by DFG; in cooperation with Carl von Ossietzky University of Oldenburg and OFFIS

Short description: Automated driving, interaction concept, impaired driving

Contact: Bianca Biebl

Projektzeitraum: 08/2020 - 05/2023


Despite the increasing degree of autonomy in vehicles, their technical functionality still requires the involvement of a human driver for supervision, support and intervention. The design of such human cyber-physical systems (HCPS) must take into account the human, the technical system and their interaction to ensure safe and efficient cooperation. In the area of sensor technology for example, strategies for safe automated vehicle guidance when detection of the environment is incomplete can be derived from the compensatory behavior of visually impaired drivers. On the other hand, HCPS have the potential to support drivers with sensory impairments such as partial visual field deficits. Such deficits can be compensated for under certain conditions. The increased task demand when adopting compensatory strategies can however increase cognitive load, which in turn impacts the selection of an appropriate decision making strategy for performing demanding maneuvers (e.g., entering intersections). In order to design and implement technical systems that can compensate for these human limitations concerning perception and cognition, an appropriate level of trust in the system must be ensured.


In this project, "learning from humans" refers to the investigation of the behavior of visually impaired drivers in order to derive compensatory strategies for an incomplete detection of the environment by sensors. Additionally, we will analyze the impact of cognitive load on the selection of decision making strategies. Finally, the third focus of this research project lies on identifying measures to promote user trust in an automated system.

"Designing for people" in this project refers to the derivation of guidelines for the design of an individualized automated system that adapts to the individual and situational needs of drivers. This system can thus compensate for deficits on a perceptual or cognitive level and promote user trust. Eventually, the designed system will be validated to generate a model for the relationship between perception, cognitive workload, decision making strategies, and trust when navigating through intersections with an HCPS.


To realize these goals, studies are conducted in a driving simulator at all three participating institutions, with all project partners using a common framework of intersection scenarios and target variables (e.g., driving behavior, gaze and scanning behavior, trust questionnaires, etc.). Above that, adaptation of the experimental design will allow the consideration of the individual research focus of all project partners (e.g., execution in an MRI scanner, testing of visually impaired participants, etc.). The definition of such a common framework before conducting the studies allows a seamless merging of the results afterwards for a joint modeling of the analyzed factors.