Date of Award

December 2016

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Engineering

First Advisor

Xiao Qin

Committee Members

Alan J. Horowitz, Yue Liu, Jie Yu, Jun Zhang

Keywords

Detector Data, Queue, Real-time, Signal, Travel Time, Vehicle Trajectory

Abstract

This dissertation attempts to develop simple and direct approaches to estimate the vehicle queue length and travel time along signalized arterial links for real-time traffic operations. This dissertation is the first to demonstrate a process using vehicle trajectory data to generate detector volume, speed and time occupancy data, along with the generalized flow rate, density and space mean speed data. This approach minimizes detector over-counting and miss-counting issues. The detection zone can be of any shape or size and at any location along the trajectory. The relationships among detector volume, speed and time occupancy along signalized arterials are analyzed theoretically and experientially. If the generalized definitions of flow rate, density and space mean speed are used, the fundamental relationship, v=ds, holds valid in a signalized arterial environment. The fundamental relationship diagram plotted using field signalized arterial data has not been seen in any of the literatures reviewed.

Within the defined time-space region, the scatter diagram of the generalized density and the detector time occupancy presents a strong linear correlation. Simply converting detector volume counts within one data collection time period to use as the generalized flow rate introduces estimation errors. There are two major reasons. The first is that vehicles don’t completely cross the detector during the data collection time period. The second is that it assumes vehicles would evenly spread across the data collection time period when crossing the detection zone. Traffic flow intensity is introduced and defined within the time-space regions to provide much more accurate description of the traffic flow arrival and departure conditions.

This dissertation attempts to make improvements to the input-output technique for queue estimation along signalized links. Based on analyses of the theoretical and experiential cumulative input-output diagrams, also known as the Newell Curves, two major improvements are proposed to improve the performance of the input-output technique. The improvements take into account vehicles stop on top of detectors in the estimation, make necessary adjustments to detector vehicle counts, and introduce a reset mechanism to remove the accumulated estimation errors during a long time period. The improvements are tested using two sets of field data. One set of data are 10-second queue and virtual detector data generated using the Federal Highway Administration Next Generation Simulation Peachtree Street dataset. The other set of data are field manually collected 20-second queue, and loop detector vehicle count and time occupancy data at metered on-ramps. It is concluded that both improvements help to produce estimation results far better than the original input-output technique. With adjusted detector vehicle counts, the performance of the Kalman Filter queue estimation model is also improved.

A simple conservation law approach is developed to estimate travel time along signalized arterial links. Inputs used include the traffic flow intensity at input and out detectors, plus the initial vehicle queue. The estimated travel time is tested with the field travel time data to evaluate the performance of the estimation. The developed model is also compared with the NCHRP Project 3-79 model and the Little’s Law queueing theory model. The developed model performs much better for per short interval travel time estimation.

The proposed travel time estimation approach only uses the detector volume and time occupancy data. It does not rely on signal timing data to estimate the control delay or a delay model to estimate the queueing delay. In addition, neither roadway geometry nor vehicle length data are used.

Download

Share

COinS