Event Title
Logit Regression Model to Predict Driver Left Turn Destination Lane Choice Behavior at Urban Intersections
Loading...
Faculty Mentor
Dr. Matthew Vechione
Document Type
Poster Presentation
Date of Publication
January 2021
Abstract
As autonomous vehicles become more prevalent during daily commutes, there will be an increased need to predict the driving behavior of humans behind the wheel. As this increase in prevalence will not be immediate, this human behavior prediction will be vital to minimize the number of traffic incidents between human-controlled and autonomous vehicles. One of the most unpredictable driving behaviors that can be observed is the destination lane choice of a driver turning left at an intersection. If this destination lane choice behavior can be predicted, then the introduction of autonomous vehicles can occur in a safer, more controlled manner. This research builds off of previous research by making use of two Next Generation Simulation (NGSIM) arterial street data sets in order to attempt to predict the turning behavior of human-controlled vehicles using real-world field data. This prediction can then be used in addition to the technology already in place in autonomous vehicles to decrease the risk of a collision at an intersection with a concurrently turning human-driven vehicle. Currently, there is no existing model that can accurately predict driver behavior due to the unpredictability of human-controlled vehicles. The resulting model will allow for a safer transition from human-controlled vehicles to autonomous vehicles. In addition to the safety benefits, this model could also be incorporated into popular microscopic traffic simulation tools, in order to improve the overall accuracy and efficiency of these tools.
Keywords
Driver Behavior, Autonomous Vehicles, Traffic Safety
Persistent Identifier
http://hdl.handle.net/10950/3075
Logit Regression Model to Predict Driver Left Turn Destination Lane Choice Behavior at Urban Intersections
As autonomous vehicles become more prevalent during daily commutes, there will be an increased need to predict the driving behavior of humans behind the wheel. As this increase in prevalence will not be immediate, this human behavior prediction will be vital to minimize the number of traffic incidents between human-controlled and autonomous vehicles. One of the most unpredictable driving behaviors that can be observed is the destination lane choice of a driver turning left at an intersection. If this destination lane choice behavior can be predicted, then the introduction of autonomous vehicles can occur in a safer, more controlled manner. This research builds off of previous research by making use of two Next Generation Simulation (NGSIM) arterial street data sets in order to attempt to predict the turning behavior of human-controlled vehicles using real-world field data. This prediction can then be used in addition to the technology already in place in autonomous vehicles to decrease the risk of a collision at an intersection with a concurrently turning human-driven vehicle. Currently, there is no existing model that can accurately predict driver behavior due to the unpredictability of human-controlled vehicles. The resulting model will allow for a safer transition from human-controlled vehicles to autonomous vehicles. In addition to the safety benefits, this model could also be incorporated into popular microscopic traffic simulation tools, in order to improve the overall accuracy and efficiency of these tools.