1. Evaluating the Benefits of a System-Wide Adaptive Ramp-Metering Strategy in Portland, Oregon
Soyoung Ahn1, Robert L. Bertini2, Christopher M. Monsere3, Oren Eshel4, and June H. Ross5
Empirical Comparison of German and U.S. Traffic Sensor Data and Impact on Driver Assistance Systems
Steven Hansen and Dr. Robert L. Bertini, Portland State University
Overview
The project seeks to evaluate the benefits of the System-Wide Adaptive Ramp Metering (SWARM) system implemented in the Portland Metropolitan area as compared to a pre-timed system.
The results presented here are drawn from a two-week pilot study of the southbound Oregon highway 217 corridor in June, The pilot study was part of an effort to develop a strategic design for a future regional-level study.
Results
VMT exhibited only a marginal increase during the study period of SWARM operation as compared to the pre-timed study period. However, VHT and average travel time increased under SWARM, corresponding to an increase in total delay on the freeway. Empirical evidence suggests that this increase resulted from higher metering rates at most of the on-ramps. These higher metering rates under SWARM resulted in lower travel times on several major on-ramps, indicating a tradeoff between an increase in freeway delay and lower on-ramp delays.
In the morning peak period on OR-217 southbound, a recurrent bottleneck is located between Scholls Ferry Rd. and Greenburg Rd., and the resulting queue propagates over 4–5 miles upstream. The bottleneck activates due to large inflow from the on-ramp at Scholls-Ferry Rd.
VHT and the average travel-time increased under SWARM, corresponding to an increase in total freeway delay.
Speed Contour under SWARM
Speed Contour under Pre
Speed Contour under Pre - -
timed Mode
timed Mode I I - - 5 5 72 72 nd nd
Ave
Ave
ORE
ORE - -
99W
99W
Greenburg Rd
Greenburg Rd
Scholls
Scholls
Ferry Rd
Ferry Rd
Performance
Comparison
VMT
VHT
Travel-Time
Delay (vehicle-hours)
Pre-Timed
65,871
1,337
8.8
210
SWARM
66,426
1,416
9.2
283
% Change
0.8%
6.0%
5.1%
34.7%
Hall Blvd
Hall Blvd
Denney Rd
Denney Rd
Allen Blvd
Allen Blvd B B - -
H Hwy
H Hwy
Walker Rd
Walker Rd
Barnes Rd
Barnes Rd US US - - 26 26
Communications
Portland City
Center
ORE 217 Southbound
Percentages of communication failures were below 2% at most locations with the pre-timed strategy.
Percentages under SWARM were much larger, exceeding 10% in some locations.
Steve Hansen, Dr. Robert Bertini, Portland State University
The SWARM strategy admitted consistently higher flows to the freeway.
Most on-ramp flows were slightly larger when SWARM was in operation. The increases in flow (except at Hall Blvd.) were between three and nine percent.
Objectives
The objective of this study was to quantify system-wide benefits in terms of reduced delay, emissions and fuel consumption, and safety improvements on and off the freeway due to implementation of the SWARM system.
Darker time-space regions show increases in total freeway delay under SWARM between 6:30 and 8:30 AM.
Changes in delay (vehicle -
hours) under SWARM B
H Hwy US 26 I 5
Hall Blvd
Scholls
Ferry Rd
Walker Rd
ORE
99W 72 nd
Ave
Greenburg Rd
Denney Rd
Allen Blvd
Wilshire Rd
Barnes Rd
Travel
Direction
South
ODOT is completing an upgrade of its data network to accommodate the large volume of data required by the SWARM algorithm to compute metering rates and send adjustments to the ramp metering system in real-time.
Background
Starting in 1981, the Oregon Department of Transportation (ODOT) implemented pre-timed ramp metering to manage traffic congestion during the morning and afternoon peak periods.
These moderate increases in flow decreased travel time on the Beaverton-Hillsdale Highway and Scholls-Ferry Road on-ramps, sampled once every five minutes.
The results of this research will assist ODOT in fine-tuning the deployment of the SWARM system and in reporting its benefits to decision-makers and the public.
The graphs below are from a test on 3/22/2007 of using programmable logic controllers (PLCs) to collect additional data about freeway on-ramp operation at Scholl’s Ferry Road on OR-217 Southbound.
Next Steps
The lessons learned from the pilot are being used in the design of a subsequent study that will include northbound I-205 and either northbound I-5 or northbound OR-217. Archival data does not include vehicle inflows to the ramp or metering rates and times and counting on-ramp inflows during the pilot study was labor intensive.
Electronic data collection devices – programmable logic controllers (PLCs) – will be installed on the key on-ramps in each corridor to measure vehicle flows entering and leaving the on-ramp. During the SWARM period of the study, the PLCs will also collect data about meter activation times and metering rates.
The SWARM system was first implemented in Southern California and was adapted for use in Portland. It has been implemented in stages since May 2005 and is operating on six of the seven metered freeway corridors.
Whether the increase in the total freeway delay was solely caused by the higher merging rates remains an open question since the bottleneck discharge rate could not be measured from the data. Moreover, delays could not be quantified at all on-ramps due to the limitations on data collection efforts, and hence, it was not feasible to analyze the system-wide trade-offs between the freeway and on-ramp delays.
SWARM Algorithm
SWARM computes metering rates in real-time under both a local mode with respect to conditions near each ramp and a global mode operating on an entire freeway system. The more restrictive rate is applied.
crit
Forecasted
density trend
Excess density
Actual density T
Density
Time
The densities around the bottleneck are forecasted by performing a linear regression on a set of data collected from the immediate past and applying a Kalman filtering process to capture non-linearity. The excess density, the difference between the forecasted density and a threshold density (saturation level at the bottleneck), is converted to the required density to avoid congestion within a forecasting time span (Tcrit).
Dashed lines represent inflows, or vehicles entering the on-ramp. Solid lines represent outflows, or vehicles entering the mainline from the on-ramp.
Acknowledgements
The Oregon Department of Transportation and the National Science Foundation provided the data and funded this research. Dennis Mitchell, Jack Marchant, Phuong Nguyen, Galen McGill and Hau Hagedorn (now with OTREC) from ODOT, Bill Kloos and Paul Zebell from the City of Portland, and Nathaniel Price from the Federal Highway Administration provided valuable input throughout this project.
1Assistant Professor, Civil and Environmental Engineering, Arizona State University,
2Associate Professor, Civil and Environmental Engineering, Portland State University,
3Assistant Professor, Civil and Environmental Engineering, Portland State University,
4Graduate Research Assistant, ITS Lab/Urban and Regional Planning, Portland State University,
5Research Coordinator, Research Unit, Oregon Department of Transportation,