ISTTT20 Tutorials

Traffic Management
and
Control

A. Delay-Based Models – Cyclic Flow Profiles

Prof. Nathan H. Gartner
University of Massachusetts
Lowell, MA, USA

© NHGartner,UML, July 2013

 Delay-Based Models – Cyclic Flow Profiles
Performance Index based on delay, stops or combination
thereof.

Most popular model:

TRANSYT

TRAffic Network StudY Tool

Developed by TRRL, GB
 TRANSYT
TRANSYT, the traffic network study tool, is a computer model
to optimize traffic signal timings and perform traffic signal
simulation.

TRANSYT has two main elements – the traffic model which is

used to calculate the performance index for a given set of
signal timings and an optimizing process that makes changes
to the settings and determines whether they improve the
performance index or not.
Structure of TRANSYT program
Network data, Optimization
Flow data data

New
Signal
Settings

Initial Optimum
Signal Signal
Settings Traffic Model Optimization Procedure Settings

Performance
Index

TRANSYT program
Delays and Graphs of
Stops in Cyclic flow
Network Profiles
SIMULATION MODEL:
 The simulation model used in TRANSYT-7F is a macroscopic
traffic simulation model.
 Platoons of vehicles are considered rather than individual
vehicles.
 TRANSYT-7F simulates traffic flow in a step-wise manner: the
cycle length is divided into small, equal time increments, called
steps.
 The average flow pattern past a point in the road network is
represented by a histogram also known as Cyclic Flow Profile.
 Three traffic flow patterns are developed:
the "IN", "GO", and "OUT" patterns.
TRANSYT Flow Patterns
“IN” Pattern
Inflow
Flow Rate (vph)

Red Green Y

SLT

“OUT” Pattern
Outflow

“GO” Pattern

Saturation Flow Rate

Time (steps)
The IN-Pattern or Arrival Flow Pattern:

The IN-Pattern is the traffic arriving at the stop line. The arrival flow is
expressed mathematically as follows:
n
IN it   Fij ( Pij .OUT jt ' )
j

where,
INit = the IN-pattern on link i for time step t
Fij = smoothing process related to platoon dispersion for flow to
link i from link j
Pij = proportion of feeding link OUT-pattern that feeds the
subject link
OUTjt' = the OUT-pattern of link j for step t'
t' = time step t minus the travel time for flow to link i from link j
n = number of links (j) that feed link I

The IN-pattern is computed for each step, t, in the cycle, thus forming a
flow profile.
The GO-Pattern or Saturation Flow Pattern:

The GO-pattern is the flow rate at each step that would leave the stop line when the traffic
signal turns green if there were enough traffic to saturate the green.

The OUT-Pattern or Departure Flow Pattern:

The OUT-pattern is the profile of traffic actually leaving the stop line. It is usually equal to the
GO-pattern as long as there is a queue. After queue dissipates, it is equal to the IN-pattern for
remainder of effective green.

The queue (or the number of vehicles held at the stop line during any time interval, t)
determines the OUT-pattern

mt  max{( mt 1  qt  st ),0}
where,
mt = number of vehicles in the queue in time interval t on a given link
(and similarly for mt-1)
qt = number of vehicles arriving in interval t, given by the IN-pattern
st = number of vehicles allowed to leave in interval t, given by GO-pattern

The OUT-pattern is given for link i during time interval t:

OUTit  mi,t 1  qit  mit

Platoon Dispersion
The start-and-stop operation of signals tends to create platoons of vehicles that travel
along a link. TRANSYT models dispersion of platoons as they progress along a link.

v ' ( t  T )  F .vt  [(1  F ).v(' t  T 1)

where,
v'(t+βT) = predicted flow rate in time interval t+βT
ofpredicted platoon
vt = flow rate of initial platoon during step t
β = factor for vehicles traveling faster than
speed limit
T = cruise travel time on link, in steps
F = smoothing factor, where:

F  (1   . .T ) 1
where,
α = empirically derived constant, platoon dispersion factor (PDF), takes
into account site-specific factors such as grade, curvature, parking,
opposing flow interference, other sources of impedance.
OPTIMIZATION MODEL:
TRANSYT uses a Performance Index (PI) which allows the user to define preference of
performance of the traffic network. Develops signal timing plan that produces an
optimal value of PI.

DI only 
PROS only 
 
PROS & DI 
 
PROS/DI 
PI   Average back of queue on link i 
DI. 
 Queuing capacity on link i 
Throughput only 
 
Throughput & 1/DI 
Throughput /DI 
 

In general, TRANSYT-7F always attempts to maximize the PI, unless the disutility index
(DI) has been selected as the PI. Since the DI can be a combination of delay, stops,
queuing, and fuel consumption, this value must be minimized.
Disutility Index

The DI is a combination of vehicle delay, stops, and fuel consumption.

The "standard" delay and stops DI optimization objective function is defined as

follows:
n
DI  {( wdi di  KwSi Si )  QP}
i 1

where,
DI = disutility index, analogous to the original TRANSYT performance index
di = delay on link i (of n links) and on an optional user-specified upstream input
link i-1
K = user-coded stop penalty factor to express the importance of stops relative to
delay
Si = stops on link i per second
wxi = link-specific weighting factors for delay (d) and stops (s) on link i
Ui = binary variable that is '1' if link-to-link weighting has been established, zero
otherwise
QP = queuing penalty
Queuing Penalty

The queuing penalty is used to minimize the possibility of spillback.

QP  QBiWq (qi  qci ) 2

where,
Q = a binary variable,
'1' if the queue penalty is included in the DI
'0' otherwise
Bi = a binary variable,
'1' if the maximum back of queue (qi) exceeds the queuing
capacity
'0' otherwise
Wq = a network-wide penalty applied to the excess queue spillback
qi = computed maximum back of queue on link i
qci = queuing capacity for link i
PROS

 Progression opportunities (PROS) represent ability of vehicles to

progress through multiple intersections without stopping. Points are
scored every time a vehicle is able to traverse additional intersections.
Therefore it is possible to achieve PROS without having a wide
bandwidth stretching from the beginning of an arterial street to the
end.
 It is also possible to select a combination of PROS and DI in order to
achieve a compromise between the two objectives. For example, if the
PI is defined as PROS/DI, the program attempts to simultaneously
maximize the numerator (PROS) and minimize the denominator (DI).
 For optimizing bandwidth and progression speeds, PROS/DI is
recommended.
 PROS-only is rarely acceptable, because although it often produces a
good bandwidth, the minor movements are allowed to fail.
 Example

10%
TM

1000
vph 10% TM
TRANSYT generated arrival flow profiles

Intersection - 2 Intersection - 3

Intersection - 4 Intersection - 5
Periodic delay-offset relationship

Delay vs. Offset Curve 1500 vph, 30%TM

40.00

35.00

30.00
Averafe delay, s

25.00

20.00

15.00

10.00

5.00

0.00
0 50 100 150 200
offset, s
 TRANSYT – Current Versions

US: TRANSYT-7F
http://mctrans.ce.ufl.edu/featured/transyt-7f/

UK: TRANSYT 14
https://www.trlsoftware.co.uk/products/junction_signal_design/transyt/
 ISTTT20 Tutorials

Traffic Management
and
Control

B. Progression Models – Arterial Streets

Prof. Nathan H. Gartner
University of Massachusetts
Lowell, MA, USA

© NHGartner,UML, July 2013

Advantages of Coordination of Traffic Lights

•Higher level of traffic service: higher overall speeds and

reduced number of stops.

•Traffic flows more smoothly, often with an improvement in

capacity due to reduced headways.

•Vehicle speeds are more uniform: no incentive to travel at

excessively high speeds to reach a green light; slow drivers
are encouraged to speed up to avoid having to stop for a
red light.
 Advantages (cont’d)

•Fewer accidents because platoons arrive at each signal

when it is green, reducing red signal violations and
rear-end collisions.
•Greater obedience to signal commands by motorists
and pedestrians.
•Through traffic tends to stay on the arterial streets
instead of diverting onto parallel minor streets in
search of alternative routes.
 Approaches for Arterial Signal Timings

• Maximize the bandwidth

of the progression
or
• Minimize overall delays and stops

• We want to combine both.

 Traffic Progression Methods

• Bandwidth optimization
• Robust solution of the traffic control problem
• Optimal phase sequencing
• Advance queue clearance
• Progression speed adjustment
Basic Bandwidth Maximization Problem

• objective function :

Maximize
b  k b
Time Space Diagram for MILP-1
 Directional Interference Constraints

• Progression bands use only the available green time

and do not infringe upon red times.

wi  bi  1  ri
w i  bi  1  ri
 Arterial-Loop Integer Constraint
If we proceed along a loop consisting of the following points:
•Center of inbound red at Si
•Center of outbound red at Si
•Center of inbound red at Sh
•Center of outbound red at Sh,
we end up at a point that is removed an integral number of cycle
times from the point of departure:

and hi  hi   h   i  hi

rh  rh ri  ri
t hi  t hi   ( wh  wh )   ( wi  wi )  ( i   h )   h   i  hi
2 2
• Cycle time constraint:
1 / C2  z  1 / C1

• Progression speed constraints:

di d di d
z  ti  i z and z  ti  i
fi ei fi ei

di d d and di d d
z  i t i 1  t i  i z z  i t i 1  t i  i z
hi di 1 gi hi di 1 gi
 w
CANAL STREET
NEW ORLEANS, LA. [552]
S N
k = 1:1.2
(0,432,120) (0,1576,0)
DISTANCE
(0,417,0) (0,297,216) [513] (FEET)
E
1
k1
1:3.1 NODE1 371
(0,432,0) (0,1360,0)

2
[584]
(0,420,128) (0,744,0)
NODE2 371
1:3.0 (0,504,56) (0,1360,324)

Example: (0,156,0)
3
(328,0,600) [928]

1:1.3 NODE3 390

Data for Canal St.
(0,672,160) (0,1084,0)
(0,458,0) (0,298,110)
[408]
4
[942]
(0,942,0) (0,964,0)
New Orleans, LA 1:1.5
NODE4
(0,672,0)
(0,215,0)
(0,974,22)
(0,215,272) [487]
276

5
[383] (0,259,124) (0,327,0)
NODE5 472
(0,796,0) (0,724,68)
1:1
(0,48,0) (0,48,8)
[ 56]
6
[244] (0,0,244) (0,0,0)
NODE6 364
1.3:1 (0,1040,0) (0,784,0)

[239] 7
(12,27,200) (0,27,0)
1.6:1 NODE7 968
(0,1223,17) (0,772,0)
(0,764,0) (122,747,0)
[869]
[1026] 8
(0,875,151) (0,919,0)
NODE8 377
1.8:1 (0,1436,60) (0,772,44)
(0,224,0) (0,164,16) [180]
[16] 9
(0,0,16) (0,0,0)
NODE9
(0,1452,0) (0,800,0)

[800]
 MILP-1
Canal St.
Symmetric Progression
 MILP-1
Canal St.
Uniform Weighted Progression
Weight: Total Volume Ratio
Variable Bandwidth
 The Variable Bandwidth Problem
• Different bandwidth for each directional road
section of the arterial
• Individually weighted with respect to its
contribution to the objective function
• Width can vary and adapt to the prevailing traffic
volumes on each link
• User can still choose a uniform bandwidth
progression if desired.
 Directional Interference Constraints

The time reference point at each signal is redefined to

the centerline of the band (progression line), rather
than the edge.
• Outbound direction:
bi b bi b
 wi  ( 1  ri )  i  wi 1  ( 1  ri 1 )  i
2 2 2 2

• Inbound direction

bi b bi b
 wi  ( 1  ri )  i  wi 1  ( 1  ri 1 )  i
2 2 2 2
 Phase Sequencing
Pattern 1: Outbound left leads; Inbound left lags

Inbound
lij
Outbound
l ij

Pattern 2: Outbound left lags; Inbound left leads

Inbound
lij
Outbound
l ij

Pattern 3: Outbound & Inbound lefts lead

Inbound
lij OR
Outbound
l ij

Pattern 4: Outbound & Inbound lefts lag

Inbound
lij OR
Outbound
l ij
 Objective Function for MULTIBAND

Objective Function
Maximize:
1 n

n  1 i 1
ai  bi  ai  bi

with
p
 qi 
p
 qi 
ai    ai   
 si   si 
 MILP-1
Canal St.
Variable Weighted Progression
Weight: (Tot. Volume/Capacity)2
 MILP-1
Canal St.
Variable Weighted Progression
Weight: (Tot. Volume/Capacity)4
Simulation Results, Canal St.
 ISTTT20 Tutorials

Traffic Management
and
Control

C. Progression Models – Networks

Prof. Nathan H. Gartner
University of Massachusetts
Lowell, MA, USA

© NHGartner,UML, July 2013

 Network Problem
MILP-2 can be extended for controlling traffic flow in a
network of intersecting arterials.
Significantly more challenging problem:
Progressions must be provided on all the arterials of
the network, simultaneously. The principal difficulty is
due to the requirement that the progression bands of
intersecting arterials cannot overlap in time at any
junction of the network.
 Network Loop Constraint Geometry

Si,a a Si+1,a
Sl+1,d Sj,b

d b

Sl,d Sj+1,b
Sk+1,c c Sk,c
 Network Loop-Integer Constraints
For any closed loops of the network consisting of more
than 2 links, the summation of internode and intranode
offsets around a loop of intersecting arterials must be
an integer multiple of the cycle time:

( ia ),( i 1 ,a )  ( i 1 ,a ),( jb )  ( jb ),( j 1 ,b )  ( j 1 ,b ),( kc ) 

( kc ),( k 1 ,c )  ( k 1 ,c ),( ld )  ( ld ),( l 1 ,d )  ( ld ),( ia )  N
 MILP-3 Solutions
(Ann Arbor, MI Network)
Uniform & variable bandwidths
4th St., Ann Arbor network
 MILP-3 Characteristics

• Mathematical Programming problem (MILP)

• Optimization of offsets, splits, cycle length and
phase sequencing
• Single arterial and grid network cases
• Uniform or variable bandwidth optimization
• Computationally demanding solutions
 Improved Computational Procedures
Just as important as more accurate traffic
models.
Improvements enable the following:
• use more economical mathematical
programming codes which are more affordable
and more easily accessible;
• optimize larger-scale networks than would
otherwise be possible;
 Improved Computational Procedures
• assure more reliable convergence to optimal
solutions, i.e., the procedure would not fail as often;
• analyze a larger number of alternatives at a much
reduced cost (one should note that the progression
schemes may involve a multitude of data sets, choice of
coefficients and parameter ranges); and
• have the opportunity to use the codes in real time,
e.g., in a multilevel RT-TRACS system.
 Network Decomposition Approach
• A priority arterial sub-network consisting of an arterial tree is
selected from the original traffic network based on its
geometry and on the traffic volumes that each link is carrying.
• The priority sub-network is optimized first and the results are
then used for the solution of the entire network.
• Alternative sub-networks can be selected in a heuristic
procedure if further improvements are desired.
 Advantages of Network Decomposition

• Arterial tree sub-network selected based on geometry and

link volumes
• Based on traffic engineering criteria
• No relaxation of integrality of any variable in each sub-
problem
• Intermediate solutions obtained at each step consist a feasible
and optimal solution for the priority sub-network considered
• Optimal solution is obtained in the majority of cases during
the first iteration.
 Tree Selection
Priority sub-network contains only a “tree” of arterials,
eliminating any network loops.
Arterials in “tree” should include principal arterials in
network and are chosen on the following criteria:
• Choose principal arterial of the network to be trunk of
tree; include only crossing arterials in sub-network.
• Tree consists of maximum number of arterial two-way
links without forming any network loops.
Ann-Arbor, MI and Memphis, TN networks.
 Conclusions
•Mathematical programming models for the
development of optimal arterial-based progression
schemes in urban signal networks.
•Provides continuous green band in each direction
along the artery at the desired speed of travel to
•Facilitates movement of principal through flows
along the arterials