SHRP2 Naturalistic Data Analysis of Older Drivers’ Gap-Acceptance Behaviour
Open Access
Article
Conference Proceedings
Authors: Sneha Srinivasan Rammanoharan, Jose Alguindigue, Apurva Narayan, Siby Samuel
Abstract: Drivers aged 65 and older are very prone to motor vehicle crashes. Intersections appear to be hazardous for drivers of this age group due to the driver’s cognitive, perceptual, and psychomotor challenges. Literature notes that older drivers find it incredibly challenging to safely navigate left turns at signalized intersections. Studies have identified the driver’s physical health, vision, and cognition as factors that impact the ability of older drivers to sufficiently monitor the gaps in oncoming traffic to make a left turn safely. The current paper aims to address the gap in the literature by explicitly examining older drivers’ gap acceptance behaviors during left turns at protected intersections. We utilize the Naturalistic Driving Study Data collected via the Strategic Highway Research Plan (SHRP2) to understand older driver behavior better. SHRP2 makes available a geo-spatially linked, comprehensive database over a multi-year period from over 3400 participants across six sites. SHRP2 databases contain a relatively more significant proportion of younger and older drivers than the national driver population databases. This dataset includes a trip summary, vehicle data, driver questionnaire, and test battery data specifying driving history, physical and psychological conditions, demographics and exit interview data, time-series data of the drivers approaching the intersections or just after the intersections, and forward video data of the drivers approaching the intersections or just after the intersections. Data is analyzed for participants over the age of 65 and participants between the ages of 30-50. Several hundred baseline, near-crash, and crash events are obtained for comparison. The video data is annotated using the DREAM methodology. The Roadway Information Database (RID) also considers additional variables such as crash histories and traffic and weather conditions. The samples of the forward video data provide the start time and end time of each gap accepted or rejected by the turning driver, especially when turning left, during unprotected phases, and help understand the participant’s interactions with other vehicles just before and after the intersections. As the data has been collected over multiple years across multiple sites, the dataset is considered a multivariate time series model. As there is more than a one-time dependent variable, the data was analyzed using Extreme Gradient Boost (XGBoost), Long-Short Term Memory (LSTM), and Seasonal Auto-Regressive Integrated Moving Average with eXogenous factors (SARIMAX) models. These models are expected to achieve an accuracy of around 80 percent at four-way intersections and approximately 60 percent in T-intersections. We anticipate that the older drivers will exhibit longer gap acceptance times and a greater frequency of gap rejections than their younger counterparts while turning left across traffic at signalized intersections. The findings of the current study will have implications for older driver safety. Researchers may use the findings to understand gap acceptance behaviors further, while policymakers may utilize the results to design mobility guidelines.
Keywords: Older drivers, Naturalistic Data, Time series, SHRP2, gap acceptance
DOI: 10.54941/ahfe1002478
Cite this paper:
Downloads
176
Visits
411