1. June 24, 2013 Presented to SigSub
Shared Permitted Left-Turn Lane Operations Results from Comparative Testing
June 24, 2013
Presented to SigSub

2. The Issue
A follow up evaluation of the HCM procedure for evaluating left-turn operation when opposed by a shared permitted left-turn lane at a signalized intersection revealed the following:
“The HCM 2010 and HCM 2000 procedures tend to predict notably different capacity and delay at higher volumes for the one-lane SPLT and the two-lane SPLT. A detailed examination of the two models indicates that the HCM 2010 procedure predicts a higher delay and lower capacity than the HCM 2000.”
- Bonneson, J. Working Paper 2: Supplement to Working Paper 1 on Shared Permitted Left-Turn Lane Operation. November 19, 2012.

3. A Proposed Solution (4.4 Beta)
As described in Working Paper 2: Supplement to Working Paper 1 on Shared Permitted Left-Turn Lane Operation, and as presented during the January 2013 SigSub Subcommittee Meeting, the Through Car Equivalent Factor (EL2), is proposed to be modified to include the proportion of right-turn vehicles on the opposing approach, when it is determined that the presence of a right-turn vehicle on the opposing approach creates a gap for left turns on the subject approach.
Approach Configuration
Proposed Modification to EL2
One-Lane SPLT Approach
One-Lane Approach Plus a Left-Turn Bay
Two-Lane SPLT Approach
In addition: nq is proposed to be modified to reflect a sat flow rate in the opposing lane equal to 1200 veh/h/ln for one lane SPLT configurations and 1800 veh/hr/ln for through or through/right lane configurations

4. A Charge for Additional Analysis
During the January 2013 Annual meeting, SigSub decided not to adopt the “beta” fix and requested additional analysis and scenario testing against a wider range of right-turn proportions and geometric configurations, and by using a wider range of simulation software. A group of volunteers agreed to meet and develop a testing plan, execute the testing plan, and report back to SigSub with results.

5. Test Configurations
Single Lane Configurations
Multi-Lane Configurations
Geometry A
Geometry D
Geometry B
Geometry E
Geometry C
Geometry F
Performance Measure = Subject Approach Control Delay

6. Testbed Assumptions Base Saturation Flow Rate 1900 pc/h/ln Left Turns
10% of total volume
Right Turns
0, 10, 20 % of total volume
Pedestrian Volumes
0 ped/h per crosswalk
Peak Hour Factor
1.0
Lane Width
12 feet
Percent Heavy Vehicles 0%
Number of Parking Movements
0 maneuvers/h
Signal Timing Parameters
G = 39 seconds Y
4 seconds R
2 seconds e
Arrival Type 3
Area Type
Non-CBD
Other
Control is Pretimed
All approaches have same geometry
Subject Approach = EB

7. Test Software HCM 2000 (Scott Reinbacher)
HCM 2010 (Christopher Kinzel and Daniel Morgan)
Comp Engine 4.4 (w/o proposed fix)
Comp Engine Beta 6 (w/ proposed fix)
TransModeler (Daniel Morgan)
Vissim (Bill Cisco)

8. Test Results: HCS 2000, 2010 & Beta
Beta 6 version more closely matches HCS 2000 compared to 2010 v4.4 for A, B, and D
All results similar for C
Major differences in delay under E and F occur in the “protected” regime
RT proportion factor in Beta 6 influences approach delay most heavily for A & D
RT pattern “flips” under different configurations
(2010)

9. Test Results: TransModeler
TransModeler predicts higher delays than 2000 and 2010 under A, C, and F
TransModeler closely resembles HCM 2010 under D
TransModeler predicts lower delay than HCM 2000 and 2010 under E (at high LT vols)
Results highly unstable at high congestion levels

10. Test Results: Vissim
Vissim results fall in between HCM 2000 and 2010 approaching congestion under D
Vissim predicts lower delays than 2000 and 2010 under E

11. Summary of Findings from Task Force Members
Beta version more closely matches HCS 2000
Particularly when “protected” regime is ignored
But is HCS2000 the right “truth” set?
Is PRT an appropriate factor to introduce into EL2?
Could test alternative fixes to EL2 that ONLY consider the opposing lane saturation flow, not the right-turn percentage
For Configurations D and E, could adjust ssl3 rather than stretching the EL2 concept to include multi-lane opposing approaches
Redefine EL2 based only on gdiff?
Future testing should:
Test other EL2 models
Incorporate field data (ideally)
Include CORSIM results
Test different volume regimes in which all approaches are not equal
New research is needed to fully address left-turn modeling issues