Lane change maneuver quantification on a freeway based on vehicle reidentification

Traffic congestion and associated delays have become a serious problem over much of the world. To mitigate traffic congestion, it is essential to better understand the factors that cause traffic delays. It has long been recognized that Lane Change Maneuvers (LCMs) are a critical factor in traffic flow theory, and LCMs could be a contributing factor to traffic congestion. However, research on LCMs has been limited by the fact that there are no efficient methods to collect the number of LCMs in the field. The collection of LCM data currently requires labor-intensive efforts to extract the information from film or video. Image processing technologies are starting to help in this task, but for obtaining accurate LCM data the labor demands remain high. To meet the need for LCM data, this work develops an approach for LCM quantification. Specifically, this approach estimates the number of vehicles entering a lane (Nen) and the number of vehicles exiting a lane (Nex) separately. This approach is compatible with existing vehicle detectors, and it only requires data collected at two detector stations to estimate the number of LCMs between them.The proposed approach for LCM quantification employs recent advances in Vehicle Reidentification (VRI), a process to match a vehicle observation at one detector station to an observation of the same vehicle at another station. Building off of previous studies, this work develops a more robust VRI algorithm that is compatible with conventional loop detectors. This VRI algorithm is tested over several highway links. The test results show that this VRI algorithm is able to reidentify long vehicles even when the traffic conditions change between free flow and congestion.The VRI results yield the difference of Nen and Nex between a pair of consecutive reidentified vehicles. The VRI results can also be used to estimate the lower bounds and upper bounds on Nen and Nex. Thus, the difference between Nen and Nex is determined, and the values of Nen and Nex are constrained to lie between their lower bounds and upper bounds. Based on these conditions, an approach to estimate Nen and Nex is developed and three variants are proposed. A vehicle trajectory data set is used to evaluate the performance of the proposed approach for LCM quantification, since vehicle trajectory data are one of the few sources that could provide ground truth LCM information. The data set does not include loop detector data that can be used for VRI. Therefore, the VRI results are simulated to be consistent with the empirical performance of the proposed VRI algorithm. The evaluation results show that the proposed approach for LCM quantification looks promising for estimating Nen and Nex, although further testing on additional data sets is necessary.The approach for LCM quantification could eventually be used to estimate the number of LCMs from conventional loop detector data, thereby providing new insight into travel patterns between lanes and the resulting impacts.