direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Page Content

Neuroadaptive cognitive modeling

Neuroadaptive cognitive models are studied together with Oliver Klaproth and in cooperation with Airbus, Airbus Defence & Space and Zander Laboratories. The objective of is to explore means of early detection of potentially safety-critical pilot behaviour during flight.


The Out-Of-The-Loop Performance Problem

High levels of automation in human machine interaction can lead to the "Out-Of-The-Loop" performance problem. Operators are more and more forced into the role of supervisory controllers of automation; at the same time, they are expected to maintain the necessary situational awareness for assuming manual control in case of critical situations. Therefore, it is essential that the information provided by the automation is processed correctly by the user and is not overlooked and overheard even in situations of extremely high or low workload.

New feedback channel

User commands and input allow the automation to detect if and how the user conveys previously communicated information and alerts. In monitoring tasks where user input is largely absent, the automation is deprived of such feedback and it has to assume that the user’s situational awareness is perfect. Physiological measurements, especially event-based electroencephalography (EEG), provide interesting insights into whether and how the user reacts to new information. These data can be made available to the automation using a passive brain-computer interface.

Identification of safety-critical performance

Early anticipation of critical performance requires predictions of user behavior. Cognitive modeling allows for detailed simulation of the user even in dynamic task environments. By a continuous integration of data on user input, system state, and the environment, the automation maintains an updated model of its operator that enables it to generate predictions of probable behavior depending on the situational factors. The aim of the user modelling is to allow the automation to simulate the effects of its actions on the user and to provide assistance as soon as a safety-critical drop in user performance is predicted.

Neuroadaptive cognitive models

Conventional user models use data on previous ("historical") user behavior in combination with information about the state of the system and environment to generate new behavioral predictions. Such models are often unable to take into account the user's subjective view of the interaction. Neuroadaptive models, on the other hand, use physiological data to adapt to the user's mental state in order to make individual predictions for specific users. However, the greatest added value of the method lies in its diagnostic function: the combination of historical and psychophysiological user data together with information about the context allows not only the determination but also the verification of assumptions about the causes of deviant performance.

Use case cockpit

Due to increasing automation, risks of the "Out-Of-The-Loop" problem also play an important role in the aircraft cockpit. This is closely connected to the prominent issue of “automation surprise”, where automation behavior in the cockpit violates pilot expectations and can even lead to a loss of control. In simulator studies in cooperation with Airbus and Airbus Defence & Space, we investigate possible benefits of neuroadaptive cognitive modelling as the basis for intelligent cognitive systems for assistance in the cockpit



Project-related publications

Krol, L.R., Klaproth, O.W., Vernaleken, C., Wetzel, I., Gaertner, J., Russwinkel, N., & Zander, T.O. (2018, July). Towards a neuroadaptive cockpit: first results. Poster session presented at the 3rd International Mobile Brain/Body Imaging Conference Berlin, Germany.

Zusatzinformationen / Extras

Quick Access:

Schnellnavigation zur Seite über Nummerneingabe

Auxiliary Functions