Driver-state monitor for highly automated driving
Christopher will develop a system that is able to monitor the operator’s vigilance and mode awareness (i.e., awareness of automation status and activation during automated driving on highways. His work will be carried out using empirical measures based on non-intrusive eye-tracking measurements from volunteers in driving simulators.
Additionally, based on indicators such as eye-scanning behaviour, Joel and Alberto will develop a monitoring system that can estimate seconds ahead of time whether a driver is at risk for a collision under high automation. This research proposal focuses on highways, which is where highly automated driving will be implemented first. However, in the farther future, highly automated driving will also be feasible in more complex scenarios. Hence, Joel will focus on highway driving, whereas Alberto will focus on human hazard perception during (highly automated) driving in less predictable situations involving multiple road users and pedestrians.
In conjunction with Christopher, Matt will adopt a cognitive modelling approach (based on the COSMODRIVE simulation model) to capture driver perception and behaviour during automated driving and during transitions between manual and automated driving. The core topic of this modelling approach will concern the impact of automation on drivers’ mode awareness and visual scanning. In conjunction with Daniel from Zhenji from WP1, Matt will collect data regarding situations where the functional limitations of the automation are reached, such as a situation where the vehicle approaches a complex traffic situation that the computer cannot handle, and which requires switching back to manual control. Matt will also use the cognitive model as a reference for real-time detection of deviant driver behaviour.
The results of this work package will be used to improve the HMI of WP2, such that the HMI takes into account the state of the operator in real time. WP3 will yield a demonstrator (implemented together with WP2) that is able to monitor, based on eye-gaze patterns and other behavioural parameters, whether the driver is available, attentive, and aware of the situation.