FREEWAYS AND ARTERIALS TURNING MOVEMENT COUNTS ESTIMATION AND PREDICTION

Simple item page
dc.contributor.advisorHaghani, Alien_US
dc.contributor.authorNohekhan, Amiren_US
dc.contributor.departmentCivil Engineeringen_US
dc.contributor.publisherDigital Repository at the University of Marylanden_US
dc.contributor.publisherUniversity of Maryland (College Park, Md.)en_US
dc.date.accessioned2022-06-15T05:41:09Z
dc.date.available2022-06-15T05:41:09Z
dc.date.issued2022en_US
dc.description.abstractAccurate turning movement counts in interchanges and intersections are priceless information in traffic management and traffic signal design. The time-varying nature of traffic conditions can be captured with the massive deployment of traffic sensors, no longer a viable option considering the limited budget of transportation authorities. The present study proposes a framework to employ the available traffic data to estimate the turning movement counts in freeway interchanges and arterial intersections. The proposed framework for interchange turning movement count estimation illustrates that obtaining acceptable estimates of off-ramp hourly traffic volume is possible using only two days of data collection on each interchange. Next, this study investigates the intersection turning movement count estimation. The study explores this estimation from three aspects using the turning movement count data in Austin, TX. First is the model structure, for which four different machine learning models are examined. The results indicated that the multi-layer perceptron trained on all intersections and fine-tuned over each target intersection yields the best results. Second, since the traffic volume of each leg of an intersection is not always available, a two-step framework is proposed to estimate the approach volumes in the first step and then input them into the turning movement estimation model in the second step. The third aspect is the sensitivity analysis of turning movement and approach traffic counts ground-truth data sizes on the accuracy of the proposed two-step framework. These analyses indicated that collecting only five days of turning movement counts and deploying continuous traffic count sensors on a quarter of intersection approaches generate acceptable results with a median absolute error of approximately eight vehicles per 15 minutes. The application of the proposed framework in the prediction of turning movement counts reveals that accurate counts can be generated up to 30 minutes in advance. Additionally, the framework's application in traffic signal design illustrates that a single intersection's annual user delay cost can be reduced from 8 to 2.5 million dollars.en_US
dc.identifierhttps://doi.org/10.13016/ziid-oetv
dc.identifier.urihttp://hdl.handle.net/1903/28758
dc.language.isoenen_US
dc.subject.pqcontrolledCivil engineeringen_US
dc.subject.pqcontrolledTransportationen_US
dc.subject.pquncontrolledDetector deployment strategiesen_US
dc.subject.pquncontrolledInterchangesen_US
dc.subject.pquncontrolledIntersectionsen_US
dc.subject.pquncontrolledMLPen_US
dc.subject.pquncontrolledTraffic volume predictionen_US
dc.subject.pquncontrolledTurning movement counts estimationen_US
dc.titleFREEWAYS AND ARTERIALS TURNING MOVEMENT COUNTS ESTIMATION AND PREDICTIONen_US
dc.typeDissertationen_US

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Nohekhan_umd_0117E_22398.pdf
Size:
2.42 MB
Format:
Adobe Portable Document Format
Download

Collections

UMD Theses and Dissertations
Civil & Environmental Engineering Theses and Dissertations