Menja Scheer

Research Group:

Supervisor:

My research seeks to identify the real-time physiological measurements that correspond to a user's psychological states during closed-loop control tasks. This is primarily studied both in basic experiments as well as in the context of a high-fidelity flight simulator. In the course of my PhD, I intend to derive a robust online metric for measuring mental workload. The availability of reliable measurements for the actual level of mental workload will allow control machine systems to adapt to the real-time abilities and limitations of the user.

Introduction

The human operator can perform at acceptable levels even if he is experiencing high levels of workload. Nonetheless, sustained periods of high workload can cause fatigue and sudden drops in performance. Being able to assess the level of workload during an ongoing task might provide a method to predict this drops in performance.

Goals

In a closed-loop control task, such as driving, various factors can affect operational workload.  These factors include the visualization of the information, external disturbances and the complexity of the controller dynamics.

The traditional way to measure workload is by assessing the subjectively perceived workload with questionnaires, like the NASA-TLX.

To measure operator workload continuous and unobtrusive, event-related potentials, derived from the EEG, can be used instead. It has been shown that the P3, a component of the event-related potential, elicited by events in a secondary task, are sensitive to changes in primary task difficulty.

The disadvantage of this method is that a secondary task might interfere with the primary steering task. To avoid that, ignored events should be used instead, to elicit the P3. The amplitude of the P3 response to task-irrelevant stimuli is claimed to reflect the availability of general attentional resources. Thus, high workload conditions are expected to reduce its amplitude.

In the first part of this project, I investigated the feasibility of estimating operator workload for different influencing variables, with the P3, elicited by ignored sounds. Following this, I will employ online methods to classify different levels of workload during real-time closed-loop control task across different implementation scenarios in the field of aviation.

Methods

Participants are typically required to perform a basic compensatory control task in a fixed-base flight simulator, wherein they have to counteract unpredictable movements in the roll axis. Task difficulty is influenced by either modulating the bandwidth of the forcing function or by changing the order of the controller dynamics of the system. This is expected to increase operator workload. In addition, the visualization complexity of the flight task is also manipulated. As a baseline measure, workload is estimated from self-reported questionnaires (i.e., NASA-TLX).

In addition, task-irrelevant auditory stimuli are occasionally presented and their elicited P3 potentials are evaluated according to my manipulations in the control task.

Initial results

1)    Increasing the disturbance that the participants need to counteract for as well as increasing the complexity of the controller dynamics result in decreases in control performance and subjectively perceived workload. P3 amplitudes are shown to be sensitive to large changes in workload.

2)    Increasing visualization complexity diminishes our ability to accurately perceive control errors. However, it does not influence subjective workload and P3 amplitude.

Initial conclusion

The P3, elicited by ignored events, allow for different aspects of a closed-loop control task to be evaluated for their impact on the operator's workload, without disrupting the ongoing task. To do so robustly, it is necessary to develop effective measurement paradigms and identify EEG metrics that are sensitive to changes in the operator's workload.