Date of Award

Summer 2021

Publication Type

Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

First Advisor

F. Biondi

Second Advisor

B. Balasingam

Third Advisor

J. Cort

Keywords

Cognitive load estimation, Automated Driving Systems, Machine learning algorithm

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Abstract

Automation has become indispensable in all walks of everyday life. In driving environments, Automated Driving Systems (ADS) aid the driver by reducing the required workload and by improving road safety. However, the present-day ADS requires the human driver to remain vigilant at all times and be ready to take over whenever the driving task requires. Thus, continuous monitoring of the drivers is important for adopting ADS. Such monitoring can be done in ADS by measuring the cognitive load experienced by the drivers. Studies show various methods to estimate cognitive load, however, the state of the art in cognitive load estimation, particularly, the non-invasive ones suitable for ADS, still suffer from significant deficiencies. Thus, more research to improve the accuracy of cognitive load estimators is crucial for allowing the safe adoption of ADS. This thesis contains the analysis of non-invasive metrics that can be used as reliable indicators of cognitive load. Eye-tracking measures such as pupil size, eye-gaze, and eye-blinks from low-cost eye-trackers are analyzed. In addition to eye-tracking data, heart rate is also studied as an estimator of cognitive load. Furthermore, this thesis introduces a novel model-based approach to filter noisy physiological measurements for the real-time monitoring of cognitive load. The proposed measures will be beneficial to the development of more accurate metrics for cognitive load estimation, thereby contributing to the advancement of ADS. The thesis also contains a detailed description of two datasets collected at the HSLab.These datasets will be helpful to researchers interested in employing machine learning algorithms to develop predictive models of humans for applications in human-machine automation.

Share

COinS