240 likes | 353 Views
Using Archived Data to Measure Operational Benefits of a System-wide Adaptive Ramp Metering (SWARM) System Analysis of Archived 2005 Data for OR-217 SB. Comparison before and after SWARM activation on OR-217 in November 2005, and with results from June 2006 pilot study of SWARM on OR-217.
E N D
Using Archived Data to Measure Operational Benefits of a System-wide Adaptive Ramp Metering (SWARM) SystemAnalysis of Archived 2005 Data for OR-217 SB Comparison before and after SWARM activation on OR-217 in November 2005, and with results from June 2006 pilot study of SWARM on OR-217. 5/1/2007 ITS Lab
Objectives • Validate OR-217S SWARM pilot study results using 2005 archival data before and after SWARM implementation. • Process 2005 archival data for OR-217S • Loop detector, incident and weather data from PORTAL Archive • Reprocess 2006 pilot data to ensure comparable results. • Incorporate TAC suggestions for data presentation • Ramps and mainline • Ramp volume vs. mainline speed/flow • VHT: Stacked bar comparing mainline and ramp • Travel Time • Break out performance measures by mainline and ramp • Other • Data Quality
Study Periods Pilot Study in June, 2006 • 1 week each pre-timed and SWARM • Pre-Timed: 6/19 – 6/23/06 • Wed 6/21 excluded • SWARM: 6/26 – 6/30/06 • Fri 6/30 excluded Study of 2005 archive data • Before and after SWARM activation on 11/3/2005 • Pre-SWARM: 10/10 – 10/21/05 (2 weeks) • 10/24 – 10/28 (e.g. comm loss ~ 10% on 10/24 & 10/28 @ 72nd) • Post-SWARM: 11/7 – 11/17/05 (2 weeks) • 11/7 – 1 hour comm loss – 10% missing timestamps • Used 11/14 data • 11/21 – 11/25: Thanksgiving
Limitations • Potential sources of variation • Incidents • Depends on accuracy of Portal archived incident data • Climate • Seasonal driving patterns (e.g. school) • Missing data points interpolated within each day. • Greater communications failures than in pilot study • Time period selected to avoid most significant issues • Multi-week study period provides additional data • Can replace one “bad” day with a comparable day
Communications Failures - 2005 • Bottom chart shows problems at isolated ramps on three days pre-SWARM. • Higher percentage of failures in general and system-wide on three particular days post-SWARM.
Communications Failures - 2006 • 72nd Ave excluded from bottom chart • Capacity issues should now be addressed • Data network upgrade in late 2006.
Incidents - 2005 • Highlighted incidents fall within weeks selected for study. • Criteria? • Number of lanes affected, duration (>30 minutes) • 10/11 @ Canyon
Weather Nov 2005 Post-SWARM Oct 2005 Pre-SWARM
Adjustments • Adjusted for incidents, communication loss where most significant. • Pre-Timed • Did not adjust: • 10/11 Walker (Incident) • 10/12 72nd (Comm Loss) • 10/20-21 Walker (Comm Loss) • SWARM • Replaced data for all ramps: • Mon 11/7 with Mon 11/14 • Thu 11/17 with Thu 11/10 • Fri 11/18 with Fri 11/11 • Did not adjust ramp inflow calculations, charts • Also interpolated for missing data points
Performance Measures (6-9/10 AM) • 2006 original analysis was 6-9 am. • 2006 data reprocessed to ensure comparable time range and methods. • 2005 data • Pre-Timed VHT, Delay • Exclusions of incidents could account for shift in numbers • Wednesdays typically bigger travel days than Fridays • Numbers for 2006 data without exclusion of Wed 6/23 and Fri 6/30
VMT • VMT increased by 5.6% from the two weeks studied before SWARM activation and after. • 3.6% between four week periods before and after activation • Possible over-estimation due to communication errors and interpolation. 2005 Pre/Post SWARM June 2006 Pilot
VHT • Higher VHT in 2005 period, especially Tue – Thu • In 2005, VHT lower pre-SWARM than post-SWARM • Versus 2006 pilot study where VHT was lower pre-timed than SWARM 6/06 Pre-timed and SWARM 2005 Pre/Post SWARM
Oscillations 6/06 Pre-timed and SWARM 2005 Pre/Post SWARM
2005 Pre/Post SWARM Ramp Flow – by on-ramp • Flow at each ramp roughly comparable between 2005 and 2006. • 2005: slight decrease under SWARM • 2006: slight increase under SWARM • Note: 2006 chart does not include Greenburg, 99W, and 72nd June 2006 Pilot
Ramp Volume - Oblique • Higher on-ramp inflows during 2005 pre-SWARM period than 2005 post-SWARM or 2006 pilot • Change in behavior? • Do specific ramps account for the increase? • Implies that in 2005, SWARM was more restrictive than pre-timed from 7:30 to 8:30 am • Under SWARM ramp inflows roughly comparable between 2005 and 2006 • Slightly higher during June 2006 pilot 2005 Pre/Post SWARM June 2006 Pilot • Note: Greenburg, 72nd, and 99W are not included in 2005 chart, since they were not included in the 2006 pilot study.
Metering Activation by Ramp, 2006 June 2006 Pilot: BH Hwy June 2006 Pilot: Scholl’s Ferry • Reduced metering window under SWARM • Work in progress – show mainline
Delay – by ramp • High pre-SWARM delay for many ramps in 2005 • Particularly BH Hwy, Scholl’s, and Greenburg • Based on 2005 data, SWARM reduced delay at many ramps • But contradicts 2006 pilot study 6/06 Pre-timed and SWARM 2005 Pre/Post SWARM
Delay – by day of week • High pre-SWARM delay for many ramps in 2005 • Tue, Wed, Thu > Mon > Fri 2005 Pre/Post SWARM 6/06 Pre-timed and SWARM
Combined presentation of previous two slides Focus on BH Hwy, Scholl’s, and Greenburg Delay – 2005By Ramp and Day of Week
On-Ramp Travel Times • Not in scope of analysis of 2005 data • Travel times for BH Hwy and Scholls Ferry on-ramps • SWARM, June 2006, 23% lower at BH Hwy and 37% lower at Scholl’s Ferry • 5 minute samples from video, 6-19-06 to 6-29-06 June 2006 Pilot: BH Hwy June 2006 Pilot: Scholl’s Ferry FIGURE 9(a) Travel time on the Beaverton-Hillsdale Highway on-ramp. FIGURE 9(b) Travel time on the Scholls-Ferry Rd. on-ramp.
Summary of Findings • Increase in pre-timed VHT, Delay in 2005 vs. 2006 pilot study • ??? • Additional weeks allowed bad data to be replaced • Some effect but not a major change
Work in Progress • TAC Suggestions – Plan to incorporate as many as possible by next meeting • On ramp volume vs. mainline speed and flow • Comparison of travel time change on mainline and ramp • Comparison of VHT on ramp and mainline • Delay by time • Break up performance measures by ramp and mainline