Quantifying Drivers Foot Movements and Pedal Misapplication Errors

Thesis (Ph.D.)--University of Washington, 2015-12Pedal errors refer to the situation when the driver mistakenly presses the wrong pedal or does not press the pedal at all. A negative outcome of pedal misapplications is a sudden acceleration event, which has been associated with crashes. However, there is currently little information on the specific contributions. The goal of this dissertation is to identify the factors associated with a higher likelihood of pedal errors through models of driver’s foot movements. Data from 87 unique participants were collected from three studies: a driving simulator, parking lot study, and naturalistic driving. There were different foot trajectories observed that could be classified as a direct hit, hesitation, corrected trajectory, or pedal error. Within the pedal errors, four different sub-categories were observed: wrong pedal, slip, miss, and both pedals. These errors (3.27%) were not as common as the other foot trajectories and were therefore placed into one group for further analysis. Using a repeated logit model, pedal errors were shown to be associated with age-related and situational factors, including the location of the triggered event. Further exploration of the driver-related differences in movements was conducted using a functional principal components analysis that showed that the largest contribution to pedal errors were observed early in the foot movement, when compared to the direct hit and corrected trajectory. Exploration of the situational context was further examined using a naturalistic study, which showed that turning maneuvers were less likely associated with errors as drivers had their foot on the brake pedal more often. The parking and start-up sequence also had an impact on the likelihood of a pedal error. Hence, a tight parking maneuver with a sudden event was also used. No pedal errors were observed in this study. But there were more braking events observed in the parking study when compared to a similar event in the simulator study. In summary, the series of study showed that an algorithm could be designed to detect a potential pedal error early in the foot movement process such that an alert could be provided to driver in a reasonable timeframe to allow correction of the movement. This dissertation describes the factors that can be considered for such an algorithm, and the process to identify these factors