Predicting Drivers' Pedal Responses Using Multinomial Logit Models with Repeated Measures

• Activity Number: 23
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Government Statistics Section
• Author(s): Linda Boyle and Yuqing Wu*+ and Daniel McGehee and Kazu Ebu and James Foley
• Companies: University of Washington and University of Washington and University of Iowa and Collaborative Safety Research Center, TEMA and Collaborative Safety Research Center, TEMA
• Keywords: Driver Behavior ; multinomial logistic regression ; transportation

Abstract

Pedal misapplications sometimes result in crashes. The two most prominent factors in pedal error crashes would appear to be pre-crash parking maneuvers and advanced driver age, which is accompanied by compromised movement control, executive function and capability for error recovery. In order to simulate fast pedal movements resulting in pedal misapplication, a driving simulator study was designed to examine foot movement behavior as participants responded to a series of changing traffic signals. The pedal responses were grouped into four categories: direct hit, uncertainty, corrected trajectory and pedal errors. The multinomial logistic regression model considering random effects in the Bayesian framework was used to predict the likelihood of one of these four pedal responses. This method, based on Markov Chain Monte Carlo (MCMC) simulation, provides a method for considering the random effects while also allowing the simultaneous computation of logistic based equations. The results show that age and the traffic phase do affect the type of pedal response and implications of this are discussed in this presentation.