Analyzing Intersection Gap Acceptance Behavior with Naturalistic Driving Data
Files
TR Number
Date
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Safety at unsignalized intersections continues to be a major concern for transportation agencies and roadway users. To improve intersection safety, this project conducted a comprehensive study of gap acceptance behaviors at unsignalized intersections using the second Strategic Highway Research Program (SHRP 2) naturalistic driving study (NDS) data. The team collected 1,170 accepted and rejected gaps/lags based on 466 NDS trips at 60 unsignalized T-intersections in Washington state and North Carolina. The project team utilized a number of data sources, including time series data measuring vehicle kinematics for the analyzed trips, forward-facing and rear-view videos for the analyzed trips, driver demographic and driving history data, the SHRP 2 Roadway Information Database, and satellite images. First, the team identified the critical gaps for a number of common scenarios using three widely accepted methods: binary logistic regression, maximum likelihood method, and probability equilibrium method. Results showed an overall critical gap of 5.3 seconds for right-turning trips and 6.2 seconds for left-turning trips. The team then went on to develop a complete understanding of the factors affecting gap acceptance decisions using logistic regression and machine learning techniques. A number of factors were identified that affect drivers’ gap acceptance decisions, including being a gap instead of a lag, presence of leading and/or following vehicles, higher volume, intersection being unskewed, and increased number of through lanes. Finally, researchers further investigated drivers’ longitudinal and lateral acceleration behaviors during turning after accepting a gap and factors affecting their turning behaviors. Overall, both left- and right-turning vehicles initially accelerated quickly after they accepted a gap, and then reduced to a lower but prolonged acceleration rate while turning to reach a desired speed. For lateral acceleration, the peak value for the left-turning profile was reached later in the turning process than for the right-turning profile.