Ph.D. Thesis Defense by Christopher Toth
School of Civil and Environmental Engineering
Ph.D. Thesis Defense Announcement
Empirical Study of the Effect of Offramp Queues on Freeway Mainline Traffic Flow
Jorge Laval, Randall Guensler
Michael Hunter, Michael Rodgers, Baabak Ashuri (ARCH)
Date & Time: Tuesday, September 9, 2013, 5:00pm
Location: SEB 122
Congestion is a problem many major cities must endure. Understanding how congestion propagates throughout a network of freeways is an essential component of mitigating congestion. The underlying driver behavior that plays a role in congestion propagation is studied. Video data collected on the I-85 corridor in Atlanta is collected and processed to analyze driver behavior. The quality and quantity of data collected using state-of-the-art object tracking techniques is unprecedented, and can be used for other traffic flow studies in the future.
A macroscopic lane-changing model is developed for vehicles changing lane from a freeway through lane to an exit lane that leads to another freeway. The relationship between the number of lane changes, the speed of the ramp lane (using tracked vehicle data), and the location upstream of the ramp split is examined. The number of lane changes is approximated by a non-homogeneous Poisson distribution, with a parameter which is a function of ramp lane speed, and location. There is a parabolic relationship between the number of lane changes and target lane speed, and the number of lane changes is gamma-distributed with respect to distance upstream of the ramp. The parameters of the gamma distribution are a function of ramp lane speed. Data were collected from a secondary site, and fit to the model to provide additional credence to the model’s validity. A discussion regarding the role of how other access points may affect the shape of the model is opened, but not analyzed in detail due to lack of available data. The macroscopic lane changing model presented in this dissertation is best characterized as the development of generalized lane-changing relationships, and provides a starting point from which more complex corridor-level models can be developed.
This study also introduces a type of unusual car-following behavior exhibited by certain lane-changing drivers. Typically found when the ramp lanes are moving at slow speeds, some lane-changing drivers will slow down causing a disruption in the initial lane. Several case studies are used to showcase certain aspects of such driver behavior. Regression analysis is used to analyze the predictability of upstream speed of the initial lane to indicate the lane-changing disruption is responsible for the lateral propagation of congestion upstream of the location of the disruption.
The lane choice of exiting vehicles is also studied. As speeds in the general purpose lanes decreases, exiting vehicles are more likely to wait longer to move into the exit ramp lanes, resulting in an increased lane changing density. Egress from an HOT lane is studied for exiting vehicles, and the general purpose lane speeds play a role in the location distribution of where vehicles exit the HOT lane to move toward the exit.
Results from this study are expected to have the greatest impact on microscopic lane-change model validation. The data-driven macroscopic lane change model presented can be compared against simulation results to assess whether simulated drivers are behaving in a realistic manner. Additionally, results reinforce the importance of maintaining a certain speed on the roadway, as well as implications for design of freeway ramps and safety issues associated with congested freeway ramps. As data collection technologies improve and data (from high-definition cameras) becomes increasingly available, this research provides the basis for the further development of larger and more elaborate lane-changing models.