Delft Haptics Lab

Delft University of Technology

MSc. Sean Scally

Abstract

By design, current FCW systems in use by the automotive industry will present warnings as late as possible while still providing enough time to brake. The primary function of the FCW system is to alert an inattentive driver to a collision possible situation and prompt a mitigating action. Adaptive FCW systems have been shown to reduce unnecessary and nuisance alarms by adapting the timing of the warning to match the individual which improves trust. In evaluating the feasibility of an adaptive warning system, we examine its use case in a particular traffic scenario. The traffic scenario in this case, ‘lead vehicle turn-off’, is examined with the intention of revealing a route to reducing unnecessary and nuisance alarms. By looking at driver inputs in the lead up to this event, certain behaviours will indicate whether the driver has initiated a mitigating avoidance action. This information can be used to adjust the timing of warnings or suppress warnings when they are not required. The following research questions are proposed: – How do drivers of HGVs act prior to lead vehicle turn-off, specifically, is it possible to detect an effort to avoid danger prior to the presentation of a FCW?- Can this information be used to support the use of an adaptive warning system which reduces unnecessary and nuisance alarms in the given traffic scenario?In a case study with 262 instances of the traffic scenario of interest, 30 events were selected for statistical comparison. The 30 event cases selected contained data from 14 subjects. The results suggest that most cases some form of pre-emptive avoidance manoeuvre could be detected long before the presentation of a warning. 93% of cases showed some form of avoidance manoeuvre prior to the warning. 87% of drivers made an effort to decelerate prior to the warning, typically initiating the avoidance manoeuvre more than 7 seconds prior to the issuance of a warning. 27% of drivers carried out some form of steering avoidance prior to the warning. Given that the timing of early avoidance manoeuvres did not seem causally linked to a change in the contextual scene information and that, it would be difficult to predict exactly when the driver is likely to carry out an avoidance manoeuvre. For this reason, adjustable warning thresholds may offer little potential for improvement in the pursuit of increasing driver acceptance. The number of avoidance manoeuvres detected prior to the warning suggest that the system could be further improved by suppressing warnings if the driver is assumed to be attentive.

MSc Thesis
Created by The Laboratory Network