session 6 ellen grumert andreas tapani
TRANSCRIPT
Comparison of reactive algorithms for
controlling of variable speed limits
Ellen Grumert
Andreas Tapani
Transport Forum, Linköping
January 8-9, 2014
Variable speed limit system
• VSLS (Variable Speed Limit System)
• Connected variable speed limit signs
• Detectors, measuring the conditions on the road such as flow and/or mean
speed
• Decision algorithm based on flow or mean speed or both
Source: Foto taken 2010 by Holger Ellgaard, publiced at www.wikipedia.org (accessed 2011-04-13)
Source: Description of MTM, Automatic Incident Detection in the Motorway Control System MTM, March 1999
Problem to investigate
• Many of the algorithms in use in real systems are based on
simple control strategies
• May not necessary reflect the flow on the road accurately
• Many of the proposed control strategies in literature have
problems with
• Computational complexity
• Uncertainty in robustness
• Tuning difficulties of parameters – many parameters or
interpretation issues
• High data demand
1 Mainstream Traffic Flow Control (Carlson et. al. 2011)
• Aim:
• Maximize throughput at potential
bottlenecks
• Avoid congestion
• Avoid capacity drop
• How?
• Find critical density (or occupancy)
(corresponding to the maximum throughput)
• Control the inflow by lowering the speed limits
- Idea
Source: Carlson et. al. (2011)
• VSLS functionality is dependent of:
• Finding critical occupancy
• Finding suitable acceleration area
• Finding suitable application area
- Algorithm design
VSLS application area Acceleration area
1 Mainstream Traffic Flow Control (Carlson et. al. 2011)
2 Specialist (SPEed Controlling AlgorIthm using
Shockwave Theory) (Hegyi et. al. 2008, 2010)
• Based on shockwave resolution
• Tuning parameters have physical interpretation
𝒒 < 𝒒𝒄
• Idea:
• Identify the traffic states – Detect chock waves
• Predict their future evolution
• Resolve the shockwave with suitable speed limits
• Lowering speed+same density lower flow
Source: Hegyi, A. and Hoogendoorn, S.P., Dynamic speed limit control to resolve shock waves on freeways – Field test results of the SPECIALIST algorithm
2 Specialist (SPEed Controlling AlgorIthm using
Shockwave Theory) (Hegyi et. al. 2008, 2010)
3 Reducing crash potential (Lee et. al. 2003, 2006)
• Log-linear model (analouge to linear regression)
• Crash potential = crash rate
• Crash potential based on crash precursors and external
control factors
• Thresholds for lowering the speed based on crash potential
• Model calibration
• Actual crash and traffic data collected from a 10-km stretch of the
Gardiner Expressway in Toronto, Canada
• 13-month period from January 1998 to January 1999
• 234 crash cases and 234 non-crash cases.
• Crash precursors:
• Temporal variation of speed: standard deviation of speed divided by
average speed (over all lanes)
• Spatial variation of speed: difference in speed between upstream
and downstream locations
• Lane changing behavior: covariance of volume difference between
upstream and downstream locations on adjacent lanes
• External factors
• Road geometry
• Peak- or off-peak pattern
3 Reducing crash potential (Lee et. al. 2003, 2006)
Microscopic traffic simulation
• How to model congestion?
• On-ramp
• Lanedrop
• Lowering speed on one section or for a few vehicles for some time
• Potential problems
• Models vs. reality
• Calibration – data
• How realistic is each senario?
- Modeling approach
Detector area
150m
Microscopic traffic simulation
• Choice: Lanedrop
• Realistic flow levels
• Modelling of capacity drop reflects reality (based on empirical
studies from litterature)
For VSLS modelling approach 1 and 2: Find critical occupancy!
- Base case modelled in SUMO
Preliminary results – Mainstream Traffic Flow
Control
Conclusions
• Careful consideration needs to be taken regarding modelling
congestion when evaluating the algorithms.
− In SUMO a lanedrop seems to be most suitable to model congestion
• Preliminary results from Mainstream Traffic Flow Control (MTFC)
shows:
− Improved results with respect to congestion (comparing basecase with MTFC).
− Higher mean speed for MTFC compared to basecase.
• Further work:
− Investigation of two more algorithms and comparisons between the three.
• Expecations
− Different algorithms might have different advantages
− The different algorithms might be more beneficial in some specific situations
(not necessary the same)