Chapter 1

Mathematical modeling
In this chapter, we give an outline of mathematical modeling in brief also we present
the mathematical model for the traffic flow based on fluid dynamics. We discuss about
the fundamental variables of traffic flow and we also present the derivation of the
macroscopic traffic flow model which is based on [2], [4].

1.1 Outline of Mathematical modeling

Mathematical modeling is an inevitable component of scientific and technical progress.
The very formulation of the problem of mathematical modeling an object leads to a
precise plan of actions. It can be conditionally split into three stages: model-algorithm-
code (refer to the diagram).

Mathematical
modeling

Objec
t

Computer Numerical
code algorithm

Fig-1.1: Diagram of mathematical modeling

At the first stage, the “equivalent” of the object is chosen reflecting its major properties
in a mathematical form – the laws, controlling it, connections peculiar to the

Mathematical modeling
4
components, and so on. The mathematical model is investigated using theoretical
methods, enabling one to obtain important preliminary knowledge about the object.
The second stage is the choice or development of the algorithm for the realization of a
model on the computer. The model is represented in a form convenient for the
application of numerical methods. The sequence of computing and logic operations is
defined, enabling us to find out the sought quantities with required accuracy. The
computing algorithms should not distort the basic properties of the model and, hence,
of the initial object; they should be economical and convenient for the considered
problems and the computer used.
At the third stage the codes are created, “translating” the model and algorithm into a
language accessible to the computer. They also have to fulfill the economy and
convenience criteria. One can call them the electronic equivalents of the investigated
objects, already suitable for direct tests on “the experimental facility” – the computer.

1.2 Mathematical model of traffic flow

The mathematical study of traffic flow, and in particular vehicular traffic flow, is done
with the aim to get a better understanding of these phenomena and to assist in
prevention of traffic congestion problems. Traffic phenomena are complex and
nonlinear, depending on the interactions of a large number of vehicles. Moreover,
vehicles do not interact simply following the laws of mechanics, but also due to the
reactions of human drivers. It is mentioned earlier that we concentrate on macroscopic
model for vehicular traffic by assuming one-lane, one direction traffic flow of highway.

The macroscopic model describing the dynamics of macroscopic variables (e.g.

density, velocity, and flow) using partial differential equations. The independent
variables of a continuous macroscopic flow model are location x and time instant t.
To introduce the dependent traffic flow variables, consider a small segment [ x, x + dx ]
of a highway referred to as “cell x ”. Macroscopic traffic flow model describes the
dynamics of the density ρ = ρ( x, t ) , the velocity V =V ( x, t ) , and the flow
q = q ( x, t ) . This model is as follows:

∂ρ ( x, t ) ∂q ( x, t )
+ =0 − − − −( A)
∂t ∂x

Mathematical modeling
5
1.3 Basic variables of macroscopic traffic model

1.3.1 Velocity Field of Traffic

Ultimately the traffic engineer is interested in how fast cars move through the traffic
grid. Every car has a speedometer, and we all want to know how long it will take to go
from the location, A to B. Certainly one of the main quantitative measures of traffic is
the speed of cars in the traffic. Consider, for the sake of argument, a one lane highway
with cars in a line moving in the same direction. Since there is no passing, and cars
cannot move through each other, the order of the cars is preserved, although they can

move at slightly different speeds. Let the velocity of car “i” be Vi . If the X-axis

coincides with the road and the position of this car is x i ( t ) at time t, then the Calculus

tells us how Vi and x i are related: Vi is the derivative with respect to time of x i .
Thus
dx i
Vi = − − − −(1.1)
dt
Any discussion of traffic on our single-lane road must deal with a collection of vehicles,

dx i
with positions x i ( t ), i =1,2,..., N and velocities Vi = , i = 1,2,..., N . The
dt
continuum approach to traffic takes the view that this collection of discrete objects
should be replaced by a “moving continuum”, a kind of fluid of vehicles. Such a fluid
has a velocity at every value of x and at every time, and so we may define a
velocity field by a function, V ( x, t ) The idea is that the variation of V ( x, t ) with
x should be on a scale of length which is large compared to the size of a typical
vehicle. Thus the value of V ( x, t ) at a certain time t * and at a certain location x * on
the road should be the velocity of cars on that particular part off the road at that time.

If we know the velocity field for our road, how do we find the movement of an individual
car? First we must specify the car. One way to do that is to choose a particular time,

say t = t 0 , and a particular location on the road, say x = x 0 , and identify a car as
being at that spot at that time. If we then want to know where this car is located at

times, t > t 0 , we must use our knowledge of the velocity field, which tells us how fast
any car is going when at position x and time t .

Mathematical modeling
6
Thus if x(t ) is the position of our car, we know that x( t 0 ) = x 0 but also that
dx
= V ( x( t ), t ) − − − −(1.2)
dt
This last equation is the crucial one, since it relates the overall velocity field to the

function x(t ) for the particular car which was located at x 0 at time t 0 .

2.2 Density of Traffic

The second basic measure of traffic in a continuum model, in addition to the velocity
field, is the traffic density. We again imagine a one-lane road with cars spread along it.
The traffic density on this road associated with a given location x and time t , is the
average number of vehicles per unit length of road at the location and time specified.
Clearly to measure a density we need a stretch of road with enough cars on it to allow
a reasonable arithmetic mean. At the same time, we want to talk about the spatial
variation of traffic density along the road, so the length over which we average should
not be too long either, or else we will be getting to the natural scale of variation of the
density.

We will use the traditional symbol for fluid density, namely ρ , for the traffic density.
Thus ρ( x, t ) is the average number of cars per unit length at the location x and time
t.
If all vehicles have length L (or else L is a good average length) and the spacing (or
average spacing) between the cars is d , then each vehicle takes up L + d units of

1
road, so that approximates vehicles will be present per unit length of road. Thus
L +d

1
the constant density of the traffic in this case is ρ = . In this case, ρ is restricted
L+d

to a certain range, 0 ≤ ρ ≤ ρmax , where ρmax is the value at which cars are bumper to
bumper.

Mathematical modeling
7
2.3 Flux of Traffic
We think of our one-lane road, now having traffic with a certain density and velocity
field. Another thing we need to think about is the common usage of the term
traffic flow . We mean by this the rate at which what this is referring to is the rate at

which cars an observer on the edge of the road, i.e., the number of cars per unit which
cross a given point on the road. We have seen the line across a road that counts
passing vehicles. This is being used to determine the traffic flow. Actually we shall
prefer to use another term: traffic flux . The flux of cars is the wane as the flow (or

better, the flow rate) of cars- i.e., the number of vehicles going by per unit time. The
traffic flux is symbolized as q ( x, t ) .

2.4 A Key equation: Flux equals velocity times density

If there are 100 cars per km on a road, and each car is going 60 K.P.H (Km per Hour),
then in one hour 60 km worth of cars will pass an observer at the side of the road, or
60 × 100 = 6000 cars per hour. This is the flux in this example, with V = 60 km per
hour and ρ =100 vehicles per km. The flux is ρV = 60 ×100 vehicles per hour (in the
units, the “km” cancel out).

We shall use the symbol q for flux. The flux is another key function which in general
will depend, like V and ρ , upon x and t , thus
q ( x, t ) = ρ( x, t )V ( x, t ) − − − −(1.3)

1.4 Derivation of the macroscopic traffic flow

Mathematical modeling
8
 model
We consider our “bare” one-lane road (no entrances or exits). If we select some stretch
of the road, between the locations x = A and x = B > A say, we know that the number
of cars found to lie between A and B at some point t will in general depend upon
the time t . If more cars flow into segment will increase, and similar if more flow out
than in, it will decrease. We can express this mathematically in terms of the flux at A
and B . Namely, the rate of change of the number of vehicles in the segment, with

respect to time, should equal the difference in flow rate or flux. If N AB ( t ) is the number
of vehicles, then
dN AB
= −q( B, t ) + q( A, t ) − − − −(1.4)
dt

On the other hand we know that N AB can be computed from the density by
integration:
B
N AB ( t ) = ∫ ρ ( x, t )dx − − − −(1.5)
A

Thus we can write our relation as

B
d
dt ∫ ρ ( t , x )dx = −q( B, t ) + q( A, t )
A
− − − −(1.6)

We say that this last result is a global conservati on law for the vehicles on the road.

Note that the signs on the right are consistent, if q( B, t ) > q( A, t ) the more cars flow

out than in, so N AB will decrease in time. We can observe from our conversation law

(1.6) that q( B, t ) − q( A, t ) whenever ρ becomes independent of time. This is

because

∂ρ
B B
d
dt ∫ ρ( x, t )dx = ∫
A A
∂t
dx − − − −(1.7)

since A, B are constants. Note that the use of partials on the right, since ρ also
depends upon x.
The global conservation law implies a local conservation law, expressing the
conservation of vehicle number on any stretch of road sufficiently long to allow us to
assign a meaningful velocity and density function. To get the local relation we use the
fundamental theorem of integral calculus:

Mathematical modeling
9
 B
∂q( x, t )
∫ dx = q( B, t ) − q( A, t ) − − − −(1.8)
A
∂x
Note that we have used a partial derivative with respect to x here, since q depends
on both x and t.

Now equations (1.6), (1.7) and (1.8) together take the shape

 ∂ρ ( x, t ) ∂q ( x, t ) 
B

∫ 
A
∂t
+
∂x 
dx = 0 − − − −(1.9)

Now this relation holds over an interval [ A, B ] . We now use the following result
(which we don’t prove here): Let f ( x ) be continuous on some closed interval [α, β]
and assume that
B

∫ f ( x ) dx = 0
A
− − − −(1.10)

for any interval AB ∈(α, β ) . Then f ( x ) = 0 for α < x < β

Using this result, equation (1.9) becomes
∂ρ ( x, t ) ∂q( x, t )
+ =0 − − − −(1.11)
∂t ∂x
This is the model which is termed as LWR (Lighthill-Whitham-Richards) model,
developed in 1955s.
Again using equation (1.3), equation (1.11) takes the form
∂ρ ∂( ρV )
+ =0 − − − −(1.12)
∂t ∂x

One may observe the two different considerations as follows:

1 Linear model

We consider all cars are moving with some constant velocity V > 0. then the
equation for vehicle conservation (1.12) may be rewritten as
∂ρ ∂ρ
+V =0 − − − −(1.13)
∂t ∂x
This is a first order linear partial differential equation (PDE) known as
linear advection equation which is in closed form.

Mathematical modeling
10
 2 Non-linear model
If the velocity is a function of density i.e., V = V ( ρ( x, t ) ) then we have the
equation (1.12) as the form:
∂ρ ∂ ( ρV ( ρ ) )
+ =0 − − − −(1.14)
∂t ∂x

This yield
∂ρ ∂ρ dV ∂ρ
+V +ρ =0
∂t ∂x dρ ∂x
∂ρ  dV  ∂ρ
⇒ + V + ρ  =0 − − − −(1.15 )
∂t  dρ 
 ∂x
Such an equation is a first order PDE which is linear in derivatives but non-linear in
unknown, termed as quasilinea r .

1.5 Specific speed-density relationship

Mathematical modeling
11
We now focus on how to solve the non-linear equation (1.15) once V ( ρ) is given. For
the moment our concern is whether or not this assumption is justified, and then what
the function V ( ρ) should be.

On a single-lane open road this assumption seems to be fairly reasonable. An isolated

car tends to have a maximum velocity of travel, either the result of speed limits or road

conditions or driver cautions, call it V max . Then for our function V ( ρ) , we should take

V ( 0 ) =V max . We know that traffic speeds tend to go down with increasing traffic

dV
density, so we should assume that < 0, ρ > 0 . Also there is surely a density,
dρ
bumper to bumper traffic say, where the speed is essentially zero. We denote this

1
density ρmax . If L is the average car length, we could take ρmax = .
L

The interpretation and construction of the speed-density relationship plays a vital role in
the macroscopi c traffic mod el . The first steady-state speed-density relation is
introduced by Greenshields, who proposed a linear relationship between speed and

 ρ 
density that is as V ( ρ ) = V max 
1 − ρ 

 max 

Various relationships were developed following the Greenshields' direction. We use

a non linear speed − density relation ( non linear function ) which is of the form

  ρ 
2

V ( ρ ) = V max 1 −    − − − −(1.16 )
  ρ max  
 

where, V max = maximum speed (free flow speed)

ρmax = maximum density (jam density)

Mathematical modeling
12
 Fig- 1.2: Non-linear velocity-density function

With the aid of equation (1.16), equation (1.3) gives a relationship for the traffic flux
(flow) as a function of density that is presented in the following form:
 ρ3 
q ( ρ ) = V max  ρ − 2 
 − − − − (1.17)
 ρ max 
  ρ 
2

Now if we put the velocity-density function V ( ρ ) = V max 1 −    of equation (1.16)

ρ
  max  
into the general non-linear model PDE (1.14), then the explicit non-linear PDE is
obtained as in the form:

∂ρ ∂   ρ 2 
+  ρ .Vmax 1 − 2   = 0 − − − −(1.18)
∂t ∂x   ρ max  

1.6 Flow-Density Relationship

Mathematical modeling
13
The speed-density relation of freeways interpreted as being linear by Greenschilds
(1935) exerts dominant influences on the relation between the flow q [vehicles / sec]
and the density ρ [vehicles / km ] . Lighthill and Whitham conjectured a flow-density

q
relation from the information that at low density values the speed V = was regarded
ρ

N
as a function of the flow and at high density values the headway ( N is the
ρ

q
number of vehicle at certain location) was regarded as a function of speed V = .
ρ
The homogeneous velocity is considered that means it does not depend on the
highway location x or at time t explicitly. It follows that V =V ( ρ) and the equation
(1.3) gives the rate of flow (flux) is a function of density that is
q ( ρ ) = ρV ( ρ ) − − − −(1.19 )

It is obvious that the maximum speed, V max is achieved at the smallest density

ρ min → 0 , then equation (1.19) gives the results

q( 0) = 0
V ( ρ max ) = 0
⇒ q ( ρ max ) = ρ max V ( ρ max ) = 0 − − − −(1.20 )

The rate of traffic flow must be positive for all values of density, 0 < ρ < ρmax must

achieve its utmost value q max anywhere in that interval. Further, the slope of the traffic
flow is
dq dV
= V ( ρ) + ρ − − − −(1.21)
dρ dρ

The flow-density relation of Lighthill and Whitham is shaped into parabolic curve, called
fundamenta l diagram of traffic flow single-regime flow-density relation, whose
maximum value of the flow shows the expected value of the capacity of a motorway
section.

Mathematical modeling
14
 Fig 1.3: Fundamental diagram of traffic flow

Since our velocity-density relationship (1.16) is non-linear. So the flow-density relation

(1.17) is parabolic and concave. The highest flow occurs when its slope vanishes, i.e.

dq  3ρ 2 
= V max 1 − 2 =0
 − − − −(1.22)
dρ  ρ max 

It gives that the maximum traffic flow is achieved at a point of the fundamental diagram

ρ max
for ρ = , and in this case, the maximum flow is
3
2
q max = V max ρmax − − − −(1.23 )
3 3

Mathematical modeling
15