Optimization of ALINEA Ramp-metering
Control Using Genetic Algorithm with
Micro-simulation
Lianyu Chu and Xu Yang
California PATH ATMS Center
University of California, Irvine
1
Overview
• Background: ALINEA
• Genetic Algorithm
• Optimization Framework
• Simulation Modeling
• Optimization Study
• Conclusion Remarks
2
Background
• ALINEA, proposed by Papageorgiou in 1990s
• A local feedback ramp-metering strategy
• Remarkably simple, highly efficient and easily
implemented
• Good performance
– Field tests
– Simulation-based studies
• Potential applications
3
Background: ALINEA
r (t )  ~
r (t  t )  K R  (O * O(t ))
Downstream detector
On-ramp detector
Queue detector
4
Background : ALINEA
• Parameter values in field tests:
– Desired occupancy O* : 0.18 -- 0.31
– KR =70, in real-world experiments
– Downstream detector location: 40 m -- 500 m
downstream
– Update cycle t: 40 seconds -- 5 minutes
5
Background: Purposes
• How to optimize ALINEA’s operational
parameters in order to maximize its
performance?
• Method:
-> Hybrid method: simulation + GA
6
Genetic Algorithm
• Mimic the the mechanics of natural selection
and evolution
• Proven to be a useful method for optimization
• Useful when there are too many parameters
to be considered
7
Optimization Framework
Time-dependent
Travel demands
PARAMICS
simulation
Performance
Measure
Ramp Metering
Controller
Loop Data
Aggregator
MOE
GA
ALINEA
ramp-metering algorithm
Parameter
Values
8
Simulation Modeling
• Study site
 Traffic direction
Culver Dr
7
6
Jeffery Dr Sand Cnyn.
5
6.21 5.74 5.55 5.01 4.03 3.86
4
3
3.31 3.04
Irvine Central Dr
SR-133
2
1
2.35 1.93 1.57 1.11 0.93 0.6
(post-mile)
9
Simulation Modeling
• Model Calibration
Loop station @ postmile 3.04 (real world)
100
100
80
80
30-sec volume
30-sec volume
Loop station @ postmile 3.04 (simulation)
60
40
20
0
60
40
20
0
0
20
40
Percent occupancy
60
80
0
20
40
60
80
Percent occupancy
10
Optimization Study
• MOE: Total vehicle travel time (TVTT)
Ni,j: total number of vehicles that actually traveled
between origin i and destination j
Di,j: travel demand from origin i to destination j for the
whole simulation time (Di,j is not equal to Ni,j because
of the randomness of the micro-simulation)
Tki,j: travel time of the kth vehicle that traveled from
origin i to destination j
Ni , j
TVTT   Di , j  ( T / N i , j )
i , j
k 1
k
i, j
11
Optimization Study
• Setup the range of calibrated parameters for
ALINEA
Parameter
Regulator KR
Desired occupancy
Update cycle of metering rate
Location of downstream detector
Range
10 ~ 300
10% ~ 40%
10~300 sec
0~600 m
12
Optimization Study
• The best, worst and average fitness values of
each generation
13
Optimization Study
• The results of optimized ALINEA parameters
Parameter
Regulator KR
Desired occupancy
Update cycle of metering rate
Location of downstream detector
Range
70~200
19~21%
30~31%
30~60 sec
120~140 m
14
Findings
• When the regulator KR, used for adjusting the
constant disturbances of the feedback control, is
within the range from 70 to 200, the metering
system is found to perform well.
• The optimal location of the downstream detector is
found to be between 120~140 meters downstream
of the on-ramp nose in our simulation study.
15
Findings
• The update cycle of the metering rate
implementation gives the best system performance
when it ranges from 30 to 60 seconds in our study.
• The desired occupancy of the downstream detector
station is found to be within two ranges, either
from 19% to 21% or around 30% to 31%. Finally,
19% to 21% is selected for its better network
reliability performance.
16
Volume
Findings
100
90
80
70
60
50
40
30
20
10
0
0
10
20
30
40
50
60
Percent occupancy
17
Conclusions
• This paper presents a hybrid GA-simulation
method to find the optimized parameter values for
the ALINEA control. This method is effective to
find the optimized parameter values.
• Practitioners can use our optimization results as a
basic operational reference if they implement
ALINEA control in the real world.
18
Conclusions
• This study shows that micro-simulation can be
used to calibrate and optimize the operational
parameters of ramp metering control. Potentially,
micro-simulation may also be used to fine-tune
parameters for various other ITS strategies.
19
Thank you
20