Traffic Flow Models

Individual Models
• Single Regime models
– Only for free flow or congested flow
• Multi Regime Model
- Two Regime Models
– Separate equations for
• Free flow
• Congested flow
- Three Regime Models
– Separate equations for
• Free flow
• Congested flow
• Transition flow
Single Regime Models
• Greenshields Model
– Assumes linear speed-density relationships
– In order to solve numerically traffic flow
fundamentals, it requires two basic parameters
• Free flow speed (uf)
• Jam Density (kj)
• Greenshields Model

At k=0 → speed is uf
At kj → speed is zero

Model Application
y-intercept → uf The Greenshields model satisfies the boundary
conditions when the density k is approaching
Slope → negative uf /kj zero as well as when the density is approaching
the jam density kj. The Greenshields model
can therefore be used for light or dense
traffic.
Remember that: 𝑞𝑞 = 𝑢𝑢� 𝑠𝑠 . 𝑘𝑘

Multiply Greenshields equation by 𝑢𝑢� 𝑠𝑠 :

𝑢𝑢𝑓𝑓 ūf
𝑢𝑢� 𝑠𝑠 2 = 𝑢𝑢𝑓𝑓 𝑢𝑢� 𝑠𝑠 − 𝑞𝑞 (2)
𝑘𝑘𝑗𝑗
𝑢𝑢𝑢𝑢
2
Differentiate (2) with respect to 𝑢𝑢� 𝑠𝑠 :
qmax
𝑢𝑢𝑓𝑓 𝑑𝑑𝑑𝑑 Thus, the space mean speed uo
2�
𝑢𝑢𝑠𝑠 = 𝑢𝑢𝑓𝑓 −
𝑘𝑘𝑗𝑗 𝑑𝑑 𝑢𝑢� 𝑠𝑠 at which the volume is
maximum is equal to half the
free mean speed.
That is:
𝑑𝑑𝑑𝑑 𝑘𝑘𝑗𝑗 𝑘𝑘𝑗𝑗 𝑘𝑘𝑗𝑗
�𝑠𝑠
= 𝑢𝑢𝑓𝑓 − 2𝑢𝑢� 𝑠𝑠 = 𝑘𝑘𝑗𝑗 − 2𝑢𝑢� 𝑠𝑠
𝑑𝑑 𝑢𝑢 𝑢𝑢𝑓𝑓 𝑢𝑢𝑓𝑓 𝑢𝑢𝑓𝑓

For maximum flow:

𝑑𝑑𝑑𝑑 𝑘𝑘𝑗𝑗 𝑢𝑢𝑓𝑓
�𝑠𝑠
=0 𝑘𝑘𝑗𝑗 = 2�
𝑢𝑢𝑠𝑠 𝑢𝑢𝑜𝑜 =
𝑑𝑑 𝑢𝑢 𝑢𝑢𝑓𝑓 2
 𝑞𝑞
In Greenshields equation, substiture 𝑢𝑢� 𝑠𝑠 by = 𝑢𝑢� 𝑠𝑠
𝑘𝑘
𝑞𝑞 𝑢𝑢𝑓𝑓
= 𝑢𝑢𝑓𝑓 − 𝑘𝑘
𝑘𝑘 𝑘𝑘𝑗𝑗
Multiply both sides by k:
𝑢𝑢𝑓𝑓 2
𝑞𝑞 = 𝑢𝑢𝑓𝑓 𝑘𝑘 − 𝑘𝑘 (3)
𝑘𝑘𝑗𝑗

Differentiate (3) with respect to k:

𝑑𝑑𝑑𝑑 𝑢𝑢𝑓𝑓
= 𝑢𝑢𝑢𝑢 − 2𝑘𝑘
𝑑𝑑𝑑𝑑 𝑘𝑘𝑗𝑗
𝑑𝑑𝑑𝑑
At maximum flow = 0
𝑑𝑑𝑑𝑑
𝑢𝑢
0 = 𝑢𝑢𝑢𝑢 − 2𝑘𝑘 𝑓𝑓
𝑘𝑘𝑘𝑘
𝑘𝑘𝑗𝑗 𝑢𝑢𝑢𝑢 𝑘𝑘𝑘𝑘
𝑘𝑘𝑜𝑜 = ; at maximum flow 𝑢𝑢𝑜𝑜 = , 𝑘𝑘𝑜𝑜 =
2 2 2
𝑢𝑢𝑓𝑓 𝑘𝑘𝑗𝑗 𝑢𝑢𝑓𝑓𝑘𝑘𝑗𝑗
𝑞𝑞𝑚𝑚𝑚𝑚𝑚𝑚 = ūsk = × =
2 2 4
𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢
Greenshields relationship= 𝑞𝑞𝑚𝑚𝑚𝑚𝑚𝑚 = 4
Single Regime Models: Greenberg
• Second regime model was proposed after
Greenshields
• Using hydrodynamic analogy, he combined
equations of motion and one-dimensional
compressive flow and derived the following
equation
𝑘𝑘𝑗𝑗
𝑢𝑢� 𝑠𝑠 = 𝑐𝑐ln
𝑘𝑘
• Disadvantage: Free flow speed is infinite
Greenberg Model
Multiply both sides by k we obtain:
𝑘𝑘𝑗𝑗
𝑢𝑢� 𝑠𝑠 𝑘𝑘 = 𝑞𝑞 = 𝑐𝑐𝑘𝑘ln
𝑘𝑘
Differentiate q with respect to k, we obtain:
𝑑𝑑𝑑𝑑 𝑘𝑘𝑗𝑗
= 𝑐𝑐ln − 𝑐𝑐 The Greenberg model satisfies the
𝑑𝑑𝑑𝑑 𝑘𝑘
For maximum flow: boundary conditions when the
𝑑𝑑𝑑𝑑 density is approaching the jam
=0 density, but it does not satisfy the
𝑑𝑑𝑑𝑑 boundary conditions when k is
Therefore: approaching zero. The Greenberg
𝑘𝑘𝑗𝑗 model is therefore useful only for
ln =1 dense traffic conditions.
𝑘𝑘0
Substituting 1 for (𝑘𝑘𝑗𝑗 / 𝑘𝑘𝑜𝑜 ) gives:
𝑢𝑢𝑜𝑜 = 𝑐𝑐
Thus, the value of c is the speed at maximum flow.
Single Regime Models: Underwood
• Proposed models as a result of traffic studies
on Merrit Parkway in Connecticut
• Interested in free flow regime as Greenberg
model was using an infinite free flow speed
• Proposed a new model
Single Regime Models: Underwood
• Requires free flow speed (easy to compute)
• Optimum density (varies depending upon
roadway type)
• Disadvantage
– Speed never reaches zero
– Jam density is infinite

In this model, speed becomes zero only when density

reaches infinity which is the drawback of this model. Hence
this cannot be used for predicting speeds at high densities.
Single Regime Models: Northwestern Univ.
or Drake Model

• Formulation related to Underwood model

• Prior knowledge on free flow speed and
optimum density (at maximum flow)
• Speed also never goes to “zero” when
density approaches jam density
• It was built by adding constants to change
the shape of the relation to that of a bell
Single Regime Model Comparisons (1)

• Density below 20 veh/mile

– Greenberg and Underwood models underestimate
speed
• Density between 20-60 veh/mile
– All models underestimate speed and capacity
Single Regime Model Comparisons (2)
• Density from 60-90 veh/mile
– All models match very well with field data
• Density over 90 veh/mile
– Greenshields model begins to deviate from field
data
• At density of 125 veh/mile
– Speed and flow approach zero
Single Regime Model Comparisons (3)
Flow
Data Set Greenshields Greenberg Underwood Northwestern
Parameter
Max. Flow 1800- 1800 1565 1590 1810
(qmax) 2000
Free-flow 50-55 57 infinity 75 49
speed (uf)
Optimum 28-38 29 23 28 30
Speed (uo)
Jam Density 185-250 125 185 infinity infinity
(kj)
Optimum 48-65 62 68 57 61
Density (ko)
Mean - 4.7 5.4 5.0 4.6
Deviation
Multiregime Models (1)
Two-regime models

• Use Underwood model for Free flow conditions

• Use Greenberg model for congested conditions

Three-regime models

• Free flow regime

• Transitional regime
• Congested flow regime
Multiregime Models (3)
• Challenge
– Determining breakeven points
• Advantage
– Provide opportunity to compare models
– Their characteristics
– Breakeven points
Summary
• Multiregime models provide considerable
improvements over single-regime models
• But both models have their respective
– Strengths
– weaknesses
• Each model is different with continuous
spectrum of observations
Calibration of Macroscopic Traffic Models

The traffic models discussed so far can be used to

determine specific characteristics, such as the speed
and density at which maximum flow occurs, and the
jam density of a facility.

•This usually involves collecting appropriate data on

the particular facility of interest and fitting the data
points obtained to a suitable model.

•The most common approach is regression analysis.

 Regression Analysis
• Regression is done by minimizing the squares of
the differences between the observed and
expected values of a dependent variable.

• When the dependent variable is linearly related

to the independent variable, the process is
known as linear regression analysis.

• When the relationship is with two or more

independent variables, the process is known as
multiple linear regression analysis.
 Regression Analysis
• Regression is done by minimizing the squares of
the differences between the observed and
expected values of a dependent variable.

• When the dependent variable is linearly related

to the independent variable, the process is
known as linear regression analysis.

• When the relationship is with two or more

independent variables, the process is known as
multiple linear regression analysis.
 Regression Analysis
• If a dependent variable 𝑦𝑦 and an independent variable 𝑥𝑥 are related by
an estimated regression function, then:
𝑦𝑦 = 𝑎𝑎 + 𝑏𝑏𝑏𝑏

• The constants a and b are determined as:

𝑛𝑛 𝑛𝑛
1 𝑏𝑏
𝑎𝑎 = � 𝑦𝑦𝑖𝑖 − � 𝑥𝑥𝑖𝑖 = 𝑦𝑦� − 𝑏𝑏𝑥𝑥̅
𝑛𝑛 𝑛𝑛
𝑖𝑖=1 𝑖𝑖=1
and
1
∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖 𝑦𝑦𝑖𝑖 − (∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖 ) ∑𝑛𝑛𝑖𝑖=1 𝑦𝑦𝑖𝑖
𝑏𝑏 = 𝑛𝑛
1
∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖2 − (∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖 )2
𝑛𝑛

where:
n = number of sets of observations
𝑥𝑥𝑖𝑖 = ith observation for 𝑥𝑥
𝑦𝑦𝑖𝑖 = ith observation for 𝑦𝑦
 Regression Analysis
A measure commonly used to determine the suitability
of an estimated regression function is the coefficient of
determination R2 (square of the estimated correlation
coefficient), which is given by:
𝑛𝑛 2
2
∑ (𝑌𝑌
𝑖𝑖=1 𝑖𝑖 − 𝑦𝑦)
�
𝑅𝑅 = 𝑛𝑛
� 2
∑𝑖𝑖=1(𝑦𝑦𝑖𝑖 − 𝑦𝑦)

where :
Yi: is the value of the dependent variables as computed
from the regression equation. The close R2 is to 1 the
better the regression fits.
Example 1: Model Calibration
Use the data in the table (columns 1& 2) to demonstrate the use of regression analysis in
fitting speed and density data to the Greenshields Model.

Speed, us, (km/h) Density, k (veh/km)

x iy i x i2
yi xi
53.2 20 1064 400
48.1 27 1298.7 729
44.8 35 1568 1225
40.1 44 1764.4 1936
37.3 52 1939.6 2704
35.2 58 2041.6 3364
34.1 60 2046 3600
27.2 64 1740.8 4096
20.4 70 1428 4900
17.5 75 1312.5 5625
14.6 82 1197.2 6724
13.1 90 1179 8100
11.2 100 1120 10000
8 115 920 13225
Σ= 404.8 Σ= 892 Σ= 20,619.8 Σ= 66,628
𝑦𝑦� = 28.91 ̅ 63.71
𝑥𝑥=
Example 1: Greenshields Model Calibration
Using the equation:
𝑥𝑥̅ = 63.71, 𝑦𝑦� = 28.91
𝑎𝑎 = 28.91 − 63.71𝑏𝑏
1
∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖 𝑦𝑦𝑖𝑖 − (∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖 ) ∑𝑛𝑛𝑖𝑖=1 𝑦𝑦𝑖𝑖
𝑏𝑏 = 𝑛𝑛
1
∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖2 − (∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖 )2
𝑛𝑛
(892)(404.8)
20,619.8 −
𝑏𝑏 = 14 = −0.53
(892) 2
66,628 −
14
𝑎𝑎 = 28.91 − 63.71 −0.53 = 62.68
𝑢𝑢𝑓𝑓
Therefore: 𝑢𝑢𝑓𝑓 = 62.68, = 0.53, 𝑘𝑘𝑗𝑗 = 118 veh/km
𝑘𝑘𝑗𝑗
𝑢𝑢𝑓𝑓 𝑘𝑘𝑗𝑗
𝑞𝑞𝑚𝑚𝑚𝑚𝑚𝑚 = = 1849 veh/h
4

𝑘𝑘𝑜𝑜 = 59 veh/km

𝑢𝑢𝑜𝑜 = 31.3 km/h

Example 1: Greenshields Model Calibration
𝑢𝑢� 𝑠𝑠 = 62.68 − 0.53𝑘𝑘
Observed Speed (𝑦𝑦𝑖𝑖 ) Greenshields Model (𝑌𝑌𝑖𝑖 ) � 2
(𝑦𝑦𝑖𝑖 −𝑦𝑦) � 2
(𝑌𝑌𝑖𝑖 −𝑦𝑦)
53.2 52.08 589.80 536.65
48.1 48.37 368.09 378.52
44.8 44.13 252.36 231.52
40.1 39.36 125.12 109.11
37.3 35.12 70.32 38.51
35.2 31.94 39.51 9.15
34.1 30.88 26.89 3.86
27.2 28.76 2.94 0.02
20.4 25.58 72.49 11.12
17.5 22.93 130.29 35.81
14.6 19.22 204.90 93.98
13.1 14.98 250.09 194.16
11.2 9.68 313.80 369.96
8 1.73 437.41 738.99
𝑦𝑦� = 28.91 Σ=2884.00 Σ=2751.38
2751.38
𝑅𝑅2 = = 0.95
2884
Example 1: Greenberg Model Calibration
Use the data in the table (columns 1& 2) to demonstrate the use of regression analysis in
fitting speed and density data to the Greenberg Model.
Speed, us Density, k
lnki (xi) xiyi xi2
(km/h) yi (veh/km)
53.2 20 2.996 159.373 8.974
48.1 27 3.296 158.530 10.863
44.8 35 3.555 159.280 12.640
40.1 44 3.784 151.746 14.320
37.3 52 3.951 147.381 15.612
35.2 58 4.06 142.928 16.487
34.1 60 4.094 139.617 16.764
27.2 64 4.159 113.122 17.296
20.4 70 4.248 86.669 18.050
17.5 75 4.317 75.556 18.641
14.6 82 4.407 64.338 19.419
13.1 90 4.5 58.948 20.248
11.2 100 4.605 51.578 21.208
8 115 4.745 37.959 22.514
Σ=404.8 Σ=56.72 Σ=1547.02 Σ=233.04
𝑦𝑦� =28.91 𝑥𝑥̅ =4.05
Example 1: Greenberg Model Calibration
𝑘𝑘𝑗𝑗
Using the equation: 𝑢𝑢� 𝑠𝑠 = 𝑐𝑐ln = 𝑐𝑐ln𝑘𝑘𝑗𝑗 − 𝑐𝑐ln𝑘𝑘
𝑘𝑘
𝑎𝑎 = 𝑐𝑐ln𝑘𝑘𝑗𝑗
𝑏𝑏 = −𝑐𝑐
𝑥𝑥 = ln𝑘𝑘
y = 𝑢𝑢� 𝑠𝑠
𝑛𝑛 𝑛𝑛
1 𝑏𝑏
𝑎𝑎 = � 𝑦𝑦𝑖𝑖 − � 𝑥𝑥𝑖𝑖 = 𝑦𝑦� − 𝑏𝑏𝑥𝑥̅
𝑛𝑛 𝑛𝑛
𝑖𝑖=1 𝑖𝑖=1

1 (56.72)(404.8)
∑𝑛𝑛 𝑛𝑛 𝑛𝑛
𝑖𝑖=1 𝑥𝑥𝑖𝑖 𝑦𝑦𝑖𝑖 −𝑛𝑛(∑𝑖𝑖=1 𝑥𝑥𝑖𝑖 ) ∑𝑖𝑖=1 𝑦𝑦𝑖𝑖 1547.02−
𝑏𝑏 = 1 = 1
14
= −28.68
∑𝑛𝑛 𝑥𝑥 2 − (∑𝑛𝑛 𝑥𝑥 )2
𝑖𝑖=1 𝑖𝑖 𝑛𝑛 𝑖𝑖=1 𝑖𝑖 233.04−14(56.72)2

𝑎𝑎 = 28.91 − (−28.68) 4.05 = 145.06

𝑐𝑐 = 28.68 (speed for maximum flow)

Example 1: Greenberg Model Calibration
Since 𝑎𝑎 = 𝑐𝑐ln𝑘𝑘𝑗𝑗
𝑎𝑎 = 145.06 = 28.68 ln𝑘𝑘𝑗𝑗
ln𝑘𝑘𝑗𝑗 = 5.06
𝑘𝑘𝑗𝑗 =157 veh/km
Given that 𝑐𝑐 = 28.68 (speed for maximum flow)
157
𝑢𝑢� 𝑠𝑠 = 28.68ln
𝑘𝑘
𝑘𝑘𝑗𝑗
At maximum flow: ln = 1
𝑘𝑘0
ln𝑘𝑘𝑗𝑗 = 1 + ln𝑘𝑘𝑜𝑜
ln𝑘𝑘𝑜𝑜 = 4.06
𝑘𝑘𝑜𝑜 = 58 veh/km

𝑞𝑞𝑚𝑚𝑚𝑚𝑚𝑚 = 58×28.68 =1663 veh/h

Example 1: Greenberg Model Calibration
28.68ln(157/k)
Observed Speed Density, k Greenberg Model
� 2
(𝑦𝑦𝑖𝑖 −𝑦𝑦) � 2
(𝑌𝑌𝑖𝑖 −𝑦𝑦)
(yi) (veh/km) (Yi)
53.2 20 59.096 589.80 910.91
48.1 27 50.489 368.09 465.45
44.8 35 43.046 252.36 199.70
40.1 44 36.483 125.12 57.28
37.3 52 31.691 70.32 7.71
35.2 58 28.560 39.51 0.13
34.1 60 27.587 26.89 1.76
27.2 64 25.736 2.94 10.10
20.4 70 23.166 72.49 33.04
17.5 75 21.188 130.29 59.70
14.6 82 18.628 204.90 105.80
13.1 90 15.959 250.09 167.85
11.2 100 12.937 313.80 255.28
8 115 8.928 437.41 399.43
Σ=404.8 Sum 2884.00 2674.13
𝑦𝑦� =28.91
2674.13
𝑅𝑅2 = = 0.93 ⇒ Greenshields is a better fit for the data.
2884