Limitations of adaptive signals

advertisement
David Hale (Leidos)
 Ineffective
heuristic methods
 Ineffective at certain congestion levels
 Lack of accountability
 Low market share
 Competing technologies
ASCT’s optimize complex networks in only a few seconds.
Is this enough time to produce a good solution?
Or, is it only enough time to produce a “quick and dirty” solution?
Industrial engineering experts would not be impressed.
 Fast




“Equisat”
Webster’s method
Greedy algorithm
Hill-climbing
 Slow



Methods (not effective)
Methods (effective)
Simulated annealing
Evolutionary algorithms
Derivative-free methods
Delay Reduction
y
z
C
B
x
A
Run Time
Non-Adaptive,
Quasi-Adaptive
Fully Adaptive
Quick Optimization
Thorough Optimization
(none)
Delay
“If geared toward sporadic demand, they’re effective. In corridors with
defined peaks and aggressive timing, they experience diminishing returns.”
0.80
Degree of Saturation
1.00
“The industry needs a tool to quantify myriad adaptive products.”
“People are implementing these systems without a real analysis.”
 Proprietary,
secret algorithms
 Capacity analysis? NO
 Simulation? DIFFICULT



Most products can’t do it
VISSIM API = extra $$$
Too much time, money, expertise
“There is just way too much marketing.”
 Advertising




Capitalism, good and bad
Cherry-picked case studies
67% said ASCT was good for oversaturated conditions
Law & Order
 ITE



over science?
Community discussion on ASCT
Experts with decades of signal experience
ASCT is just “one tool in the toolbox”
ASCT “has its place”
“With less than 5% market share after 5 decades, acceptance is not
consistent with successful technologies.”
“Adaptive control in its infancy?”
 Decades
of availability (1960’s)
 SCATS (1976), SCOOT (1981)
 Fewer than 5% of signals are adaptive
 Why?


Costs too high
Uncertainty about benefits
“Automated performance measures allow agencies to optimize and
manage signals without an adaptive system.”
 Data
driven
 Quasi-adaptive
 Stronger algorithms
“The industry needs a tool to quantify myriad adaptive products.”
“People are implementing these systems without a real analysis.”
“If geared toward sporadic demand, they’re effective. In corridors with
defined peaks and aggressive timing, they experience diminishing returns.”
“We only use them when other options have failed.”
“There is just way too much marketing.”
“Clearly the jury is out on where they should be deployed.”
“From my experience it is smoke and mirrors.”
“With less than 5% market share after 5 decades, acceptance is not
consistent with successful technologies.”
“Automated performance measures allow agencies to optimize and manage
signals without an adaptive system.”
 Traffic

Delay not sensitive below 80% saturation
 Traffic






too heavy = not cost effective
Everything is pre-timed over 120% saturation
No platoon progression
Cycle, offsets, phasing sequence insignificant
 Traffic

too light = not cost effective
medium = sometimes cost effective
Sporadic demand (movie theater, football stadium)
Unpredictable pedestrian activity
Emergency vehicles
Incidents / accidents
David Hale (Leidos)
Download