Transcript A demonstration of traffic control algorithm analysis in the field

A demonstration of
traffic control algorithm
analysis in the field
Christian Chong-White
J. Carlos Aydos
Andrew Mehaffey,
Roads and Maritime Services
A demonstration of
traffic control algorithm analysis in the field
• Field trial – executive summary
• Algorithm change
• Trial area
• Field trial – explaining analysis
• Field trial – signals
• Field trial – flow inefficiency
• Field trial – flow throughput
• Longitudinal analysis
• Field trial – waiting time
• Field trial – customer value
• Conclusion
Field trial – executive summary
WHY
SCATS Algorithm trial assessed traffic performance effect of removing biases
on core SCATS algorithm, ie “to make SCATS more pure”.
WHAT
Biases were removed with software switches.
Removed a bias that effectively promote shorter greens over longer greens.
HOW
Trial was run for 3 months in two separate metro Sydney regions then
analysed and compared to same months previous year before trial.
WHERE
Taren Point (TAR) region and Sutherland (SUTH) region, both south of
Sydney CBD.
Algorithm change
Main title
Degree of Saturation (%)
Trial assessed traffic performance with the removal of biases on
SCATS control variable favoring short traffic signal phases.
160
150
140
130
120
110
100
90
80
Current SCATS
Field trial
0
10
20
30
Green length (sec)
40
Plot show SCATS control variable versus phase length.
50
Trial area
SCATS Regions
NSYD (green) – simulation study
North Sydney
Corridor subset of SCATS region
TAR (blue) – 1st trial region
Taren point
SUTH (red) Sutherland
1. Initially, control – no trial
2. Then, 2nd trial region
Sydney
CBD
Field trial – explaining analysis
Background
(12 months)
Content
Y axis:
Average
daily performance
metric value
Trial
(3 months)
Trial start
(blue line)
Pretrial
previous year
(3 months)
X axis: day
(2 years)
Analysis delay
(2 weeks)
Field trial - signals
Traffic signal effect was to reduce daily average red and green.
But daily total red reduced, but daily total green increased.
Main title
Content
Field trial – flow inefficiency
Main title
Traffic performance flow effect was to reduce both inefficient
spare capacity (undemanded flow) and congested flow (unmet flow).
Content
Field trial – flow throughput
Main title
But there was a concerning small reduction in throughput (realised flow).
Content
Longitudinal analysis
Reduced realised flow confirmed due to background seasonal effect
and not trial.
RealisedFlow (veh/h)
Longitudinal daily average realised flow versus month
Trial happened to start
in characteristic
seasonal decreasing
trend
Months of year (1-12)
Field trial – waiting time
Traffic performance travel time effect was to reduce
waiting time at traffic signals (servicing time cost).
Content
Field trial – customer value
VALUE
FUTURE
Delivered to road customers in 2 trial regions at 90 traffic control
sites:
• Total 604 h/year travel time savings
• Valued at $20M/year
When rolled out across Sydney metro at 2814 sites, estimated
• Total 9M h/year travel time savings
• Valued at $295M/year savings
or 0.1% of annual Gross Regional Product (GRP) of Sydney metro
Conclusions
•
•
•
•
Main title
Small control algorithm changes affect signal traffic outcomes
More pure adaptive traffic control performed better
Small improvements in sum produce large economic value
Seasonal background effects can confound performance analysis
Content
Questions?
Main title
Key contact for further information
Christian Chong-White
Traffic Algorithm Development Manager
Content
[email protected]