Empirical analysis of freeway traffic oscillation: Its safety impact and evolution.

摘要

Traffic oscillations are typical features of congested traffic flow that are characterized by recurring decelerations followed by accelerations. However, people have limited knowledge on this complex topic.;This dissertation first employs a matched case-control design to investigate the impact of freeway traffic oscillations on traffic safety. Traffic conditions prior to each crash were taken as cases, while traffic conditions during the same periods on days without crashes were taken as controls. These were also matched by presence of congestion, geometry and weather. Conditional logistic regression models were developed based on the case-control samples. To verify consistency in the results, 20 different sets of controls were randomly extracted from the candidate pool for varying control-case ratios. Results reveal that the likelihood of a crash increases by about 8 percent with an additional unit increase in the standard deviation of speed. The average traffic states prior to crashes were less significant than the speed variations in congestion.;The above finding underscores importance of understanding traffic oscillation's evolution. The origins of oscillations have been identified and the microscopic features of their propagations have been measured using vehicle trajectories in the framework of wavelet transform (WT). The wavelet-based energy distributions in trajectories reveal that oscillations can originate by either lane changing maneuvers (LCM) or car-following (CF) behavior. However, their developments and propagations are similar regardless of the causal factor. Furthermore, a negative linear relationship between an oscillation's duration and tau (a parameter in Newell's CF theory) is discovered.;Because of a key role of LCM in traffic oscillation's evolution, this dissertation finally investigates the LCM's impact on a lane changer's immediate follower in the target lane. Three components (i.e., anticipation, relaxation, and a behavioral change) are systematically measured. Results show that a well-established relaxation model using passing rate measurements can be extended to capture dynamics in anticipation processes. Furthermore, LCM has a regressive effect: a timid driver tends to become less timid and an aggressive driver less aggressive. An extended model based on Newell's CF theory is developed to describe this regressive effect.