Modeling HOV lane choice behavior for microscopic simulation models and its application to evaluation of HOV lane operation strategies

1
Modeling HOV lane choice behavior for microscopic
simulation models and its application to
evaluation of HOV lane operation strategies

• Jun-Seok Oh
• Western Michigan University
• Lianyu Chu
• University of California, Irvine

2
Investigation of HOV Modeling Capability in
Microscopic simulation Models

• Jun-Seok Oh
• Western Michigan University
• Lianyu Chu
• University of California, Irvine

3
Content

• Motivation and Objectives
• Classification and Operation of HOV System
• Analytical Model for HOV Lane Traffic Estimation
• HOV Modeling in Microsimulation Models
• Experiment and Performance Comparison
• New Modeling Approach
• Concluding Remarks

4
Motivation

• FHWA encourages the installation of HOV lanes as
  an important part of an area-wide approach
• There are still questions on
• the effectiveness of HOV systems
• their impacts on air quality
• The benefits of HOV systems have not been well
  quantified
• Microsimulation might be a good way, but still
  involves some limitations

5
Objectives

• Compare HOV modeling capability and performance
  in
• Paramics
• AIMSUN
• Identify limitations and investigate methods to
  enhance HOV behavior modeling in microsimulation
• Develop an improved HOV simulation analysis tool
  using API capability

6
Classification of HOV System
Infrastructure Mainline HOV lane Freeway-to-freeway direct connectors Direct local access ramps Freeway ramp meter bypass lanes Toll plaza bypass lanes
Designed Access Open system Closed system (Limits access with barrier)
Use Restriction 2 people minimum occupancy 3 people minimum occupancy Buses Vehicles paying toll (High Occupancy Toll)
Operational Period Full time operation Part time operation
7
HOV Operations
8
Analytical Model for HOV Lane Traffic Estimation

• User Equilibrium between HOV/GP
• HOV lane is faster than GP lanes
• tHL tGL
• fHOV(VHOV - VHG) fGP(VSOV VHG)
• If fHOV(VHOV) fGP(VSOV), VHG 0
• If fHOV(VHOV) gt fGP(VSOV), VHG gt 0
• VHG can be found by solving
• fHOV(VHOV - VHG) fGP(VSOV VHG)

9
HOV Modeling in Microsimulation Models

• Vehicle Types
• SOV HOV
• Defining HOV Lane (Open HOV System)
• Allow HOV only on HOV lane
• Lane barrier (Closed HOV System)
• Closed HOV available in AIMSUN
• Closed HOV via plug-in in Paramics

10
HOV Behavior Modeling

• Optional
• By allowing HOV only on HOV lane
• May underestimate HOV on HOV lane
• Compulsory
• By forcing all HOV to use HOV lane
• 100 HOV on HOV lane ? Unrealistic
• Separate links for HOV lane
• Route choice with dynamic feedback
• Not applicable to Open HOV
• Paramics provides HOV plug-in for more HOVs on
  HOV lanes

11
Experiment Scenarios

• Scenario 1 Closed HOV
• Using given capability
• Scenario 2 Separate Links for Closed HOV
• Treating closed HOV lanes as separated links
• Scenario 3 Open HOV
• No barrier between HOVL GPL
• Assumption HOV demand - 15 of total traffic
• MOEs
• Traffic volume split between HOVL GPL
• HOV demand split b/w HOVL GPL
• HOV demand split w.r.t speed of GPL

12
Study Network I-405, Irvine, California
13
Study Network I-405, Irvine, California
HOV open
HOV closed
HOV closed
HOV open

• Northbound I-405
• 6 km freeway stretch

14
Scenario 1 Closed HOV

• Paramics Plug-in provided by vendor
• add additional layers of detail to the HOV
  modeling
• influence lane changing behavior and lane
  discipline
• model both open/closed HOV lanes
• AIMSUN Default function
• Restrict lane-changing with solid-line

15
S1 Volume Comparison

• GP lane volume
• HOV lane traffic is underestimated
• Paramics HOV lane traffic constant during
  simulation period

16
S1 HOV traffic

• of HOV lane traffic
• of HOVs on HOVL

17
S1 HOVs on HOVL w.r.t GPL Speed

• Paramics
• Not sensitive to the traffic condition on GPL
• AIMSUN
• Slower speed on GPL leads to more HOVs on HOVL

18
Scenario 2 Separate links for closed HOV lanes

• Separate links for closed HOV lanes
• Use route choice model in HOV lane choice
• Dynamic link costs update
• HOVs are treated as guided drivers
• change route (lane) while driving

19
S2 Volume Comparison

• of HOV lane traffic
• Close to observed HOVL volume
• of HOVs on HOVL
• 70 80 during congested period

20
S2 HOVs on HOVL w.r.t GPL Speed

• Paramics
• AIMSUN

21
Scenario 3 Open HOV Lane

• HOV can access anywhere
• HOV lanes are restricted only for HOVs
• Rely only on lane-changing restriction model

Dotted-line all open area
22
S3 Volume Comparison

• of HOV lane traffic
• Underestimates HOVL volume
• of HOVs on HOVL
• Low HOLV utilization

23
S3 HOVs on HOVL w.r.t GPL Speed

• Paramics
• AIMSUN

24
Findings

• Closed HOV Lanes
• Underestimates HOVL traffic
• Paramics 65, AIMSUN 85 of observed
• Paramics Plug-in need improvement
• Better when incorporating route choice behavior
  with dynamic cost update
• Performance varies by route choice model
• Open HOV Lanes
• Current HOV modeling NOT satisfactory
• Paramics 60, AIMSUN 78 of observed
• Underestimates due to the lack of capability to
  measure lane-by-lane traffic condition

25
Other Scenarios

• Compulsory HOV Lane
• AIMSUN has an option for compulsory HOV
• Almost 100 HOVs use HOVL
• Not realistic for HOV lane analysis
• Useful tool for exclusive bus-lane
• Paramics
• Can implement by defining HOV only lane and SOV
  only lane
• But need to define area where both types can use
  for exiting and entering
• No HOV Lane

26
Overall Travel Time Comparison

• Limited analyses
• Compulsory and No HOV lane case outperformed
• Elasticity of HOV demand NOT considered

27
New HOV Modeling Approach

• Using API (Applications Programming Interface)
  capability
• Consider HOV drivers visual perception on
  traffic condition
• Visual perception-based instant HOV lane choice
  model

28
Concluding Remark

• Microsimulation needs to be enhanced for HOV
  analysis
• Closed HOV can be analyzed by incorporating route
  choice model with separate HOV links
• Open HOV analysis needs enhanced model
• Need to develop improved HOV behavior model
  considering drivers visual perception on traffic
  condition
• Need to calibrate model using real-world data
• HOV demand and elasticity survey
• Microsimulation has potential for HOV evaluation,
  but only with enhanced behavior model

29
Thank you!