5.

0 STATISTICAL DISTRIBUTIONS AND QUEUING THEORY

5.1 Introduction
Statistical Distributions: In designing new traffic facilities or new control plans, it is necessary to predict
the performance of traffic with respect to some particular characteristic (e.g., the frequency of headways
of a particular size, the number of cars likely to arrive in an interval, and speeds exceeding a certain
value). It is often desirable to be able to make a prediction with a minimum amount of data available or
assumed. For instance, it may be necessary in designing a pedestrian control system to predict the
frequency of headways of greater than 10 sec; in designing a left-turn pocket it may be necessary to
predict how many times per hour the number of cars arriving during one signal cycle will exceed four.
Statistical distribution models enable the traffic engineer to make these predictions with a minimal
amount of information.

Statistical distributions are useful in describing a wide variety of phenomena where there is a high
element of randomness. In traffic the most important distributions are counting distributions-those useful
in describing the occurrence of things that can be counted-and interval distributions-those useful in
describing the occurrence of the (time) intervals between events. Distributions are also used,
however, in describing such phenomena as speeds and gap acceptance

‘Traffic’ in traffic engineering refers to vehicle, people or both that pass a specified point during a
given time period. Traffic can be human or vehicular. Traffic can be motorized or non-motorized.
Knowledge of fundamental flow characteristics and associated analytical techniques is an essential
requirement in planning, design and operation of transportation systems. Traffic flow
characteristics include;

 Time headway (h) and gap (g),

 Flow (q),
 Time-space trajectory,
 Speed (u),
 Distance headway /spacing (s) & clearance (c),
 Density (k)

Traffic flow analytical techniques include

  Supply-demand modeling,
 Capacity and level of service analysis,
 Traffic stream modeling,
 Shock- wave analysis,
 Queueing analysis,
 Simulation modeling.

Traffic engineers and planners assess traffic and environmental impacts of proposed system
modifications and thus can be accomplished only through a supply-demand framework that
requires understanding of flow characteristics and their interactions. Designers determine link
sizes and configure systems and must carefully evaluate the trade-off between traffic flow levels
and levels of service. Operators identify locations and causes of existing system defects and
generate operational improvement plans and predict their effects; all these require traffic flow
analysis.

Understanding the basic principles of traffic flow and traffic flow analysis is vital in designing and
operating transport systems with greatest possible efficiency and safety.

The analytical process for transport systems (w.r.t traffic flow) consist of predicting an output as
function of specified inputs. A flow chart is shown below to illustrate this analytical process and
to emphasize the importance of knowing fundamental flow characteristics and basic analytical
techniques.

Flow Chart for Traffic flow characteristics Analytical Process

Shock-wave analysis and queueing analysis are techniques available for the analysis of
oversaturated traffic systems. Traffic stream models serve as a beginning point for shock wave
analysis.
Shock-wave analysis
Flow-speed -density states change overs space and time. When these changes of states occur, a
boundary is established that demarks the time-space domain of the flow state from another. This
boundary is referred to as shockwave.

In some situations, the shockwave can be very mild like platoon of high-speed vehicles catching
up to a slightly slow-moving vehicle. In other situations, the shockwave can be very a significant
change in flow states, as when high speed vehicles approach a queue of stopped vehicles.

Shockwaves are defined as boundary conditions in the time-space domain that demark a
discontinuity in flow-density conditions

Queueing models

When demand exceeds capacity for a period of time or an initial time headway is less than the
service time (at microscopic level) at a specific location, a queue if formed. The queue maybe a
moving or a stopped one. Essentially excess vehicles are stopped upstream of the bottleneck or
service area, and their departure is delayed to a later time period. Queueing or waiting line
phenomena are everyday occurrences. Examples include vehicles waiting to be served at a petrol
station, passengers or vehicles lined up at a transit terminal, aero planes waiting clearance for take-
off or landing, vehicles at an intersection, toll plazas, parking facilities etc.

No matter how complex, queueing systems are characterized by;

 An arrival pattern,
 A service facility,
 A queueing discipline.

When all three components are constant, the system can be analyzed by deterministic methods.
Probabilistic systems are however more common.
Two analytical techniques can be applied in studying queueing processes;
i) Shockwave analysis,
ii) Queueing analysis
Shockwave analysis can be applied when the demand-capacity process is deterministic, and is
particularly well suited to evaluating the space occupied by the queueing process and to interacting
queueing processes.
Queuing analysis can be employed for deterministic and stochastic (probabilistic) processes, and
the vehicles in the process are considered as being stored in a vertical queue.

Input requirements and classification

Input requirements of queueing analysis include the following five elements;
i) Mean arrival time,
ii) Arrival distribution,
iii) Mean service value,
iv) Service distribution,
v) Queueing discipline.

The mean arrival value is expressed as a flow rate, such as vehicles/hour, or as time headway, such
as seconds/vehicle. The arrival distribution can be specified as a deterministic distribution or a
probabilistic distribution. The term demand or input is sometimes substituted for the term arrival.

The mean service value is expressed as a flow rate such as vehicles/hour, or as time headway such
as seconds /vehicle. The service distribution can also be specified as a deterministic or probabilistic
distribution. The term ‘capacity’, ‘departure’ or ‘output’ is sometimes substituted for the term
service.

The fifth and last element to be specified is queueing discipline. The most common queue
discipline encountered is referred to as ‘first in, first out’. That is the vehicles are served in the
order in which they arrive, and the often symbol ‘FIFO’ is used to designate this queue discipline.
Other queue disciplines include ‘first in, last out’(FILO) systems encountered in elevators and
‘served in random order’(SIRO) systems.

A classification scheme in needed to assess the input characteristic in order to select the appropriate
queuing analysis: deterministic or stochastic queueing analysis. If either the arrival and /or the
service distribution is probabilistic, the exact arrival and /or service time of each vehicle is
unknown, and stochastic queueing analysis must be selected. On the other hand, if both arrival and
service distribution are deterministic, the arrival and service times of each vehicle are known and
deterministic queueing analysis is selected. The next two major portions of the unit are directed to
deterministic queuing analysis and stochastic queueing analysis

5.2 QUEUEING THEORY

Queueing theory is a collection of mathematical models of various queueing systems. It is used

extensively to analyze production and service processes exhibiting random variability in market
demand (arrival times) and service times.

Can you tell why queues form?

Queues or waiting lines arise when demand for a service facility exceeds capacity of that facility,
i.e. customers do not get service immediately upon request but must wait, or the service facilities
stand idle and wait for customers. Some customers wait when the total number of customers
requiring service exceeds the number of service facilities, some service facilities stand idle when
the total number of service facilities exceed the number of customers requiring service.

Waiting lines or queues are a common occurrence both in everyday life or in variety of business
and industrial situations. Most waiting line problems are centered about the question of finding the
ideal level of service that a firm should provide.

For example;

 Supermarkets must decide how many cash register check out points should be opened.
 Gasoline stations must decide how many pumps should be opened and how many
attendants should be on duty
 Manufacturing plants must decide the optimal number of mechanics to have on duty in
each shift to repair machines that break down
 Banks must decide how many teller windows to keep open to serve customers during
various hours of the day.
Evolution of queueing theory

Queueing theory has its beginning in the research work of a Danish engineer named A. K. Erlang.
In 1909 Erlang experimented with fluctuating demand in telephone traffic. 8 years later he
published a report addressing the delays in automatic dialing equipment. At the end of World War
II, Erlang’s early work was extended to more general and to business applications of waiting lines.
Some examples of waiting lines/queues are given in the table below;

Situation Arriving customers Service facility

Passage of customers through a supermarket Shoppers Checkout
checkout
Flow of automobiles traffic through a road automobiles Road network
network
Manually placed assembly line Parts to be assembled Assembling line
Inventory items in a warehouse Order for withdrawal Warehouse
Banking transactions Bank patrons Bank tellers
Ships entering a port Ships Docks
Maintenance and repair of machines Machine break-down Repair crew
Scheduling of patients in a clinic Patients Medical care
Number of runways at an airport Airplanes Runways
Parking lot Automobiles Parking spaces
Capacity of a motel Motorists Lodging facilities
Arrival of automobile at a garage Automobiles Repair of automobiles
Transfer of electronic messages Electronic messages Transmission lines
Flow of computer programmes through a Computer programmes Central processing unit
computer system
Sale of theatre tickets Theatre-goers Ticket windows
Arrival of trucks at central market Trucks Loading crew
Registration of unemployed at an Unemployed Registration assistants
employment exchange personnel
Calls at a police control room Service calls Policemen

Components of a basic queueing process

Firstly, there are some basic components in every queueing system;

BASIC COMPONENTS OF A QUEUEING SYSTEM

Input Source of Queue

An input source is characterized by;

 Size of calling population

 Pattern of arrivals at the system
 Behavior of arrivals
Customers requiring a service are generated at different times by an input source, commonly
known as population. The rate at which customers arrive at a service facility is determined by the
arrival process.

Size of calling population

The size represents the total number of potential customers who will require the service.

According to source

The source of the customers can be finite or infinite. For example, all people of a city or a state (or
others) could be potential customers at a supermarket. The number of people being very large, it
can be taken to be infinite. Whereas, there are many situations in businesses and industrial
conditions where we cannot consider the population to be infinite -it is finite.
 According to numbers
The customers may arrive for a service individually of in groups. Single arrivals are illustrated by
patients waiting for a doctor, students reaching at a library counter etc. on the other hand, families
visiting restaurants, ships discharging cargo at a dock are examples of bulk or batch arrivals.
According to time
Customers arrive in the system at a service facility according to some known schedule (for example
one patient every 15 minutes or one a candidate for interview (every half an hour) or else they
arrive randomly. Arrivals are considered random when they are independent of one another and
their occurrence cannot be predicted exactly. The queueing models wherein customers’ arrivals
are known with certainty are categorized as deterministic models. (insofar as this characteristic is
concerned) and are easier to handle. On the other hand, a substantial majority of the queueing
models are based on the premise that the customers enter the system stochastically at random pints
in time.
Patterns of arrivals at the system
The arrival process (or pattern) of customers to the service system is classified into two categories;
static and dynamic. These two are further classified based on the nature of the arrival rate and the
control that can be exercised on the arrival process.
In static arrival process, the control depends on the nature of arrival rate
(random or constant). Random arrivals are either at a constant rate or varying with time. Thus, to
analyze the queueing system, it is necessary to attempt to describe the probability distributions of
arrivals. For such distributions we obtain average time between successive arrivals, also called
inter-arrival time (time between two consecutive arrivals), and the average arrival rate (i.e. number
of customers arriving per unit of time at the service system.

The dynamic arrival process is controlled by both service facility and

customers. The service facility adjusts its capacity to match changes in demand intensity, by either
varying the staffing levels at different times of service, varying service charges (such as telephone
call charges at different hours of the day or week) at different timings, or allowing entry with
appointments.
Frequently in queueing problems, the number of arrivals per unit time can be estimated by a
probability distribution known as Poisson distribution, as it adequately supports many real-world
situations.

Parameters of queuing systems

The behavior of a queuing system is thus dependent on:
 Arrival process (l and distribution of inter-arrival times)
 Service process (m and distribution of service times)
 Number of servers
 Capacity of the system
 Size of the population for this system

Output variables
 Utilization rate ρ (server utilization, percentage of the time that a server is busy, where
c=the number of parallel servers)
 Probability of n customers in the system Pn
 Average number of customers in the system L (service and queue)
 Average number of customers in the queue Lq
 Average time spent by a customer in the system w (service and queue)
 Average time spent by a customer in the queue wq

Transient versus steady-state behavior

 Transient behavior (from t=0) performance indicators such as average waiting time,
average number of customers in queue, etc. are dependent of the time, e.g. wq(t), Lq(t)
 Steady-state (stationary) behavior (t → ∞) performance indicators such as average
waiting time are not dependent of the time anymore; the probability that the system is in a
certain state is completely independent of time, e.g. wq, Lq
Transient behavior
 History of number of customers in the system = graph of number of customers versus
time (solved using Queuing Diagrams graphically for Deterministic queuing systems)
 Important to know for the queuing strategy are queuing disciplines, namely:
o FIFO (first in first out)
o LIFO (last in first out = stack)
o SIRO (service in random order)
o SPT (shortest processing time first)
o PR (priority)

Markov models – Steady-state behavior

 Arrival times are exponentially distributed (Poisson process)
 Formulas for steady-state situation
 Service times are drawn from a deterministic, exponentially, or Erlang distribution
 M/M/1, M/G/1, M/Ek/1, M/D/1
 M/M/1/N, M/M/c, M/M/1/K/K
 Formulas for e.g. Pn, L, Lq, w, wq
  Formulas for M/M/1 system can be derived quite easily (from which others equations are
derived (as will be discussed later here)

Behavior of arrivals

Another thing to consider in the queuing structure is the behavior or attitude of the customers
entering the queueing system. On this basis the customers may be classified as being;

a) Patient, or
b) Impatient
If a customer on arriving at the service system stays in the system until served, no matter how long
he has to wait for service is called a patient customer. Machines arriving at a maintenance shop
in a plant are examples of patient customers.

Whereas, the customer who wait for a certain time in the queue and leaves the service without
getting service due to certain reasons such as a long queue Infront of him is called an impatient
customer.

Below are some interesting observations of human behaviors in queues;

 Balking – some customers, even before joining the queue get discouraged by seeing the
number of customers already in the service system or estimating the excessive waiting time
for a desired service, decide to return for service at a later time. In queueing theory this is
known as balking
 Reneging – customers after joining the queue, wait for service sometime and leave the
service system due to intolerable delay, so they renege.
For example, a customer who has just arrived at a grocery store and finds that a the salesman are
busy in serving the customers already in the system, will either for the service till his/her patience
is exhausted or estimates that his waiting time maybe excessive and so leaves immediately to seek
service elsewhere.

 Jockeying – customers who switch from one queue to another hoping to receive a service
more quickly are said to be jockeying.
SERVICE SYSTEM

The service is provided by aa service facility (or facilities). This maybe a person (a bank teller, a
barber) a machine (an elevator, a gasoline pump) or a space (an airport runway, parking lot, a
hospital bed) to mention just a few. A service facility may include one person or several people
operating as a team.

There are two aspects of a service system;

a) Configuration pf the service system,

b) The speed of the service.
a) Configuration of the service system
The customers’ entry into service system depends upon the queue conditions. If at the
time of customers’ arrival, the service is idle, then the customer is served immediately.
Otherwise the customer is asked to join the queue, which can have several configurations. By
configuration of service we mean how service facilities exist. Service systems are usually
classified in terms of their number of channels, or number of servers.

i) Single Server-Single Queue – the models that involve one queue- one service station
facility are called single serve models where customers wait till the service point is
ready to take him/her for servicing. Students arriving at a library counter is an
example of singe server facility.

ii) Single Server- Several Queues -- in this type of facility there are several queues and
the customer may join any one of them but there is only one service channel as
illustrated in the figure below;
iii) Several (parallel) Servers – Single Queue – in this type of model there is more than
one server and each server provides the same type of service facility. The customers
wait in a single queue until one of the service channels is ready to take them in for
servicing.

iv) Several Servers- Several Queues—This type of model consists of several servers
where each of the serves has a different queue. Different cash counters in an electricity
office where the customers can make payment in respect of their electricity bills
provide an example of this type of model.
 v) Several facilities in a series – In this, a customer enters first station and gets a portion
of the service and then moves on to the next station, get some service then again moves
on to the next station …… and so on, and finally leaves the system having received the
compete service. For example, machining a certain steel item may consists of cutting,
turning, knurling, grinding, and packaging operations, each of which is performed by
a single serve in a series arrangement.

b) Speed of service
In a queueing system, the speed with which service is provided can be expressed in either of two
ways- as service rate or as a service time.

 The service rate describes the number of customers serviced during a particular time period
 The service times indicates the amount of time needed to service a customer
 The service rates and times are reciprocal of each other and either of them is sufficient to
indicate the capacity of the facility.
Thus, if a cashier can attend, on an average 5 customers in an hour, the service rate would be
expressed as 5 customers /hour and service time would be equal to 12 minutes /customer.

Generally, we consider the service time only. If these service times are known exactly, the problem
can be handled easily. But as generally happens. If these are different and not known with certainty,
we have to consider the distribution of service times in order to analyze the queueing system.
Generally, the queueing models are based on the assumption that the service times are
exponentially distributed about some average service time.

QUEUE CONFIGURATION

The queueing process refers to the number of queues, and their respective
lengths. The number of queues depend upon the layout of service system. Thus, there may be a
single queue or multiple queues.

Length or size of the queue depends upon the operational situation such
as;

 Physical space,
 Legal restrictions and,
 Attitude of the customers.
In certain cases, a service system is unable to accommodate more than the required number of
customers at a time. No further customers are allowed to enter until space becomes available to
accommodate new customers. Such type of situations is referred to as finite (or limited) source
queue. Examples of finite source queues are cinema halls, restaurants etc.

On the other hand, if a service system is able to accommodate any number of customers at a time,
then it is referred to as infinite (or unlimited) source. Example is in a sale department, here the
customer orders are received, there is no restriction in the number of orders that can come in, so
that a queue of any size can form.

In many other situations, when arriving customers experience long queue(s) Infront of a service
facility, they often do not enter a service facility even though additional waiting space is available.
The queue length is such cases depend upon the attitude of the customers. For example, when a
motorist finds that there are many vehicles waiting at a petrol station, in most of the cases he does
not stop at this station but seeks the service elsewhere.

QUEUE DISCIPLINE
 In the queue structure, the important thing to know is the queue discipline.
The queue discipline is the order or manner in which customers from the queue are selected for
service. There are a number of ways in which customers in the queue re served. Some of these are:

a) Static queue disciplines are based on individual customer’s status in the queue. Few of
such disciplines are;
i) If the customers are served in the order of their arrival, then this is known as first-come,
first-served (FCFS) service discipline. Prepaid taxi queue at airport where a taxi is
engaged on first-come, first-served basis is an example of this discipline.
ii) Last-come-first-served (LCFS)—Sometimes, the customers are serviced in the reverse
order of their entry so that the ones who join the last are served first. For example, assume
that letters to be typed, or the order of forms to be processed accumulate in a pile, each new
addition being put on top of them. The typist of the clerk might process these letters or the
orders by taking each new task from the top of the pile. Thus, a just arriving task would be
the next to be serviced provided that no fresh task arrived before ii is picked up. Similarly,
people who join the elevator last are first ones to leave it.
b) Dynamic queue disciplines are based on the individual customer attributes in the queue.
Few of such disciplines are;
i) Service in random order (SIRO)—under this rule customer are selected for
service at a random order, irrespective of their arrivals in the service system. In
every customer in the queue is equally likely to be selected. The time of arrival of
the customers is, therefore, of no relevance in such as a case.
ii) Priority service— Under this rule customer are grouped in priority classes on the
basis if some attributes such as service time or urgency or according to some
identifiable characteristic, and FCFS rule is used to within each class to provide
service. Treatment of VIPs in preference to other patients in a hospital is an example
of priority service.

o SPT (shortest processing time first) is one of the queue disciplines. To which
category, static or dynamic class of disciplines, does it belong?
5.3 DETERMINISTIC QUEUEING MODELS

Queueing theory fundamentals

The fundamentals of queueing theory maybe best introduced by means of an example. Consider
an airport runway being used for takeoffs only. Air traffic rules state that there can only be one
aircraft on the runway at a time. Each aircraft spends about the same time on the runway, so that
runway can process takeoffs at fairly constant rate. In the queuing theory jargon, the runway would
be the server, and the maximum rate at which it can process takeoffs, under given conditions,
would be called at its service rate. For most transportation applications, the service rate of a facility
will be identical to its capacity under a given set of conditions.

Arrivals of aircrafts waiting to take off are not constant. A plot of the cumulative number of aircraft
demanding takeoff versus time might look like figure below.

The step function in figure 1 above is called arrival function, A(t). in most transportation
applications the arrival function (or arrival curve, if the step function is approximated with a
smooth curve) will be identical with the cumulative demand for facility or system for time period
being considered

Now suppose that each aircraft takes fixed amount of time to clear the runway. Given this service
time, it is possible to plot a departure function D*(t), which shows when each aircraft clears the
runway. It is also possible to plot a plot a function D(t), which shows when each aircraft begins its
takeoff run. The departure function D(t) is usually of more interest than D* (t) , since the takeoff
itself is a necessary part of the use of the runway, and should not be considered queueing delay.
Figure below shows the departure functions D*(t) and D(t) in relation to the arrival function A(t)

The difference between A(t) and D(t) (or D*(t)) represents aircraft which have arrived at the
runway but have nit begun or completed their takeoffs. These aircrafts waiting to be served are
said to be in a queue, and the line of the aircraft (or whatever) waiting to be served is called a
queue.

From the queueing diagram shown on the figure above, it is possible to determine a number of
interesting things:

 The number of customers in queue at any given time, or queue length

 The delay of each customer, assuming the order of service is the same as the order of arrival
 The summation of delay of all customers
 Given the density of the queue (that is the amount of space occupied by each customer in
the queue) the physical length of the queue, or the amount od space required for its storage.

These densities are illustrated in the figure below.

Q(t) , the vertical distance between A(t) and D(t) at any time, represents the number of customers
in queue at that time; wherever customers are being served on a first-come, first-served basis ,w(t)
, the horizontal distance between A(t) and D(t) , represents the delay to the customer arriving at a
time t; and the area between A(t) and D(t) ,shown shaded on the diagram below , represents the
total delay ,or summation of delay to all customers .

In many applications it is appropriate to approximate step functions A(t) and D(t), with smooth
curves, and to ignore the finite service time – if indeed, a finite service time exists. In other cases,
it’s more convenient to approximate A(t) and D(t) with stepwise linear functions.

Queueing diagrams have several important properties. It is very important that students understand
these properties and be able to apply them in actual queueing problems. These are;

i) The slope of D(t) is the departure rate; the slope of A(t) is the arrival rate.
ii) The departure rate cannot exceed the service rate or capacity of the server. It may be
less.
 iii) Cumulative departures can never exceed cumulative arrivals. D(t) can never be above
A(t) in the queueing diagram.
iv) When a queue is present, the departure rate will be equal to service rate. When no queue
is present, the departure rate will be equal to arrival rate. The queue first form when
arrival rate exceeds the service rate.

Below is a graphical illustration of a queuing diagram;

Deterministic queuing model

An excellent starting point is provided by D/D/1 queueing model, which is the simlest waiting line
model. It assumes the following;

i) Deterministic arrival,
ii) Deterministic service time,
iii) Single channel server,
iv) FIFO and
v) Infinite queue length.

For a simple queueing model, an important measure is its traffic intensity whereby;

𝑚𝑒𝑎𝑛 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑎𝑟𝑟𝑖𝑣𝑎𝑙 (𝜆)

𝑇𝑟𝑎𝑓𝑓𝑖𝑐 𝑖𝑛𝑡𝑒𝑠𝑖𝑡𝑦 (𝜌) =
𝑚𝑒𝑎𝑛 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑠𝑒𝑟𝑣𝑖𝑐𝑒 (𝜇)

𝑀𝑒𝑎𝑛 𝑖𝑛𝑡𝑒𝑟 − 𝑎𝑟𝑟𝑖𝑣𝑎𝑙 𝑡𝑖𝑚𝑒 = 1⁄𝜆

 𝑀𝑒𝑎𝑛 𝑠𝑒𝑟𝑣𝑖𝑐𝑒 𝑡𝑖𝑚𝑒 = 1⁄𝜇

𝜌 = 1⁄𝜇 / 1⁄𝜆 = λ/μ

Examples;

1)

Customers arrive at a movie theatre ticket window at the rate of 10/hour and are serviced
at a constant rate of 12/hour. Describe how this system will perform.
Solution

Mean arrival rate, 𝜆 =10/hr.

Mean service rate, μ =12/hr.

Since 𝜆˂ μ, in this situation there will be no queue and no idle time.

2)

Customers arrive at the ticket counter of a local movie theatre at the rate of
240 persons /hr. at 5.30 P.M. After 10minutes, the arrival rate declines to 60persons /hr. and
continues at that level for 20 minutes. If the time required to serve each customer is 20 seconds,
describe the performance of the system.

Solution

240 𝑝𝑒𝑟𝑠𝑜𝑛/ℎ𝑟
𝜆1 = = 4 𝑝𝑒𝑟𝑠𝑜𝑛/𝑚𝑖𝑛 t ≤ 10 min
60 𝑚𝑖𝑛/ℎ𝑟

60 𝑝𝑒𝑟𝑠𝑜𝑛/ℎ𝑟
𝜆2 = = 1 𝑝𝑒𝑟𝑠𝑜𝑛/𝑚𝑖𝑛 t > 10 min
60 𝑚𝑖𝑛/ℎ𝑟

60 𝑝𝑒𝑟𝑠𝑜𝑛/ℎ𝑟
𝜇 = = 3 𝑝𝑒𝑟𝑠𝑜𝑛𝑠 /𝑚𝑖𝑛 for all t
20 𝑚𝑖𝑛/ℎ𝑟
Therefore, the number of person arrivals at a time t equals 4t for t ≤ 10 min. also, the number of
persons arrivals [40+1(t-10)] for t > 10 min and the number of persons with tickets (departures) is
3t for all t.

These equations are depicted in the figure below. Notice that when the arrival curve is above the
departure curve a queue will exist. The queue will dissipate at the time when arrival and departure
curves intersect.

⸫ 40+1(t-10) = 3t and t = 15

Thus, the queue which begun to form at 5.30 P.M., will dissipate at 5.45P.M.

The longest queue will occur at time t = 10 of 10 persons, and the longest delay will be 13.33 -
10 = 3.33 minutes.

1 1
𝑇𝑜𝑡𝑎𝑙 𝑑𝑒𝑙𝑎𝑦 = (10 ∗ 10) + (10 ∗ 5) = 75 𝑝𝑒𝑟𝑠𝑜𝑛/𝑚𝑖𝑛
2 2

Number of persons entering delay = 45

⸫ average delay per person = 75/45 = 1.667 min.

3)
Vehicles arrive at a tool booth at a rate given by λ(t) = 5.2 – 0.20t, where λ(t)
is in vehicles /min and t is in minutes. The tollbooth operator processes the vehicles at a rate of μ
(t) = 3 where μ (t) is in vehicles /minute. Determine the total delay, maximum queue length and
the time when the queue will dissipate.

APPLICATIONS OF D/D/1 MODEL

Signalized intersection example

The signalized intersection serves as one of the best examples of deterministic queueing analysis
at the macroscopic level because its relatively simple and because of our personal experience with
such queuing processes. Queueing on only one approach to at a signalized intersection having only
two signal phases for that approach is considered.

Only an undersaturated situation is considered in this example. That is in each cycle the arrive
demand is less than the capacity of the approach, no vehicles wait longer than on the cycle, and
there is no overflow from one cycle to the next. A number of simplifying assumptions are included
with this first example of deterministic queueing analysis to emphasize the concept and reduce
confusion with more complicated real-life situations. Later examples demonstrate techniques for
handling more complicated real-life situations.

The figure below provides for all input requirements needed to solve this problem.

The arrival rate ( 𝜆) is specified in vehicles /hour and is constant for the study period. The service
rate (𝜇) has tow states: zero when signal is effectively red and up to saturation flow rate (s) when
the signal is green. Note that the service rate can be equivalent to saturation flow only when a
queue is present. Otherwise, the service rate is equal to arrival rate if the signal is green. The queue
discipline is assumed to be “first in, first out” (FIFO) system.
Directly below the flow rate versus time diagram, accumulative vehicles versus time diagram is
constructed. A horizontal line (in the figure below) such as the arrival rate ( 𝜆) appearts as a sloping
line with the slope equal to the flow rate. Thus, the arrival rate goes through the origin and the
slope up to the right with a slope equal to the arrival rate.

Now the service rate in the former figure is transformed to the latter figure. During the red period
the service rate is zero. So, the service is shown as horizontal line in the lower diagram. At the
start of the green period a queue is present, and the service rate is equal to the situation flow rate
(s). As shown in the latter figure, the cumulative arrival intersects the cumulative service line
during the green period. At this point in time the queue is dissipated and the cumulative service
line overlays the cumulative arrival line until the end of green period. Then the pattern repeats
itself, with the service rate varying again from zero to saturation flow rate to arrival flow rate.

A series of identical triangles are formed, with cumulative arrival line forming the top side of
triangles and the cumulative service line forming the other side of triangles. Each triangle
represents one cycle length and can be analyzed to calculate a set f five measures of performance;

 Time duration of queue (tQ),

 Number of vehicles experiencing queue (NQ),
 Queue length (Q),
 Individual delay (d),
 Total delay (TD).

The time duration of the queue is represented by the horizontal projection of the queuing triangle.
It starts at the beginning of the red period and continues until the queue is dissipated. Its value
caries between effective red time and cycle length., and is expressed in seconds. Two measures of
performance are associated with queue time duration: time duration of queue (t Q) and percent time
queue is present (PtQ).

The equations of these two measures are:

𝜆𝑡𝑄 = 𝜇(𝑡𝑄 − 𝑟)

𝜆𝑡𝑄 = 𝜇𝑡𝑄 − 𝜇𝑟
 𝑡𝑄 (𝜇 − 𝜆) = 𝜇𝑟

𝜇𝑟
𝑡𝑄 =
𝜇−𝜆

100𝑡𝑄
𝑃𝑡𝑄 =
𝐶

Where; 𝜆 = mean arrival rate (vehicles/hour),

𝜇 = mean service rate (vehicles /hour),

tQ = time duration of queue (seconds),

r = effective red time (seconds),

C = signal cycle time (seconds),

PtQ = percent time queue is present.

The time duration of the queue is helpful in understanding the storage and de-storage of vehicles
in the queue and is also in assessing the degree of saturation of the approach to the signalized
intersection. The figure below is an extension of the former figure with additional information
displayed.

Near the middle of the second green phase tQ terminates, and the queue is dissipated. This point in
time is projected on to the figure (a) above, and three areas are denoted. The fist area (A1)
represents the number of vehicles stored during previous red phase. When the signal changes to
green, the second area (A2) begins to enlarge. The second area (A2) denotes the number of vehicles
that are de-stored. When the second area is equal to the first area, the queue is dissipated. The third
area also has some significance, for it represents how many additional vehicles could be served
during this signal cycle.

The number of vehicles experiencing queuing is represented by vertical projection of the queueing
triangle. The first vehicle experiencing the queue is the vehicle that arrives just after the signal turn
red. All vehicles arriving during the red as well as vehicles arriving during the green but before
the queue is dissipated experience the queueing processes and are forced to stop or slow down
significantly. Its value caries between 𝜆𝑟 and 𝜆𝐶 and is expressed in number of vehicles. Three
measures of performance are associated with the number of vehicles experiencing queuing:

 Number of vehicles queued (NQ),

 Number of vehicles per cycle (N),
 Percent of vehicles queued (PNQ) ,

The equations for these three measures are;

𝜆𝑡𝑄
𝑁𝑄 =
3600

𝜆𝐶
𝑁=
3600

100𝑡𝑄
𝑃𝑁𝑄 =
𝐶

Where NQ = number of vehicles queued

N= number of vehicles per cycle

PNQ = percent of vehicles queued

The queue length is represented by the vertical distance through the triangle. At the beginning of
the red, the queue length is zero and increases to its maximum value at the end of the red period.
The queue length decreases until the arrival line intersects the service line when the queue length
is equal to zero. The queue length remains equal to zero until the end of the green period when the
pattern repeats itself. Three queue measures of performance are of primary interest;

 Maximum queue length (Qm),

 Average queue length while the queue is present (𝑄̅𝑄 ),
 Average queue length (𝑄̅).

The equations of these measures are;

𝜆𝑟
𝑄𝑚 =
3600

𝑄𝑚 𝜆𝑟
𝑄̅𝑄 = =
2 7200

𝑄𝑚 𝑡𝑄
𝑄̅ =
2𝐶

Whereby Qm = maximum queue length (vehicles),

𝑄̅𝑄 = average queue length while queue is present (vehicles),

𝑄̅= average queue length (vehicles).

Individual delay is represented by the horizontal line across the triangle. The first vehicles to arrive
after the beginning of the red encounters the largest individual delay. Each vehicle arriving
thereafter experiences smaller and smaller individual delay until the queue is dissipated. Vehicle
arriving thereafter until the beginning of the next red encounter no individual delay. Three delay
measures of performance are of primary interest:

 Maximum individual delay(dM),

 Average individual delay while queue is present (𝑑̅𝑄 ),
 The average individual delay (𝑑̅ ).

The equations for the three measures are;

𝑑𝑀 = 𝑟
 𝑟
𝑑̅𝑄 =
2

𝑟𝑡𝑄
𝑑̅ =
2𝐶

Whereby; dM = maximum individual delay (seconds),

𝑑̅𝑄 = average individual delay while queue is present (seconds),

𝑑̅ = average individual delay (seconds).

The total delay cycle is represented by the cross-sectional area of the queueing diagram triangle
and is expressed in vehicle-seconds. Any of the following equations can be used to calculate the
total delay (TD) in vehicle -seconds;

𝑁𝑄 𝑟
𝑇𝐷 =
2

𝑄𝑀 𝑡𝑄
𝑇𝐷 =
2

𝑇𝐷 = 𝑑̅ 𝑁

Queueing patterns
A variety of queueing patterns can be encountered and a classification of these patterns is proposed
in this section. The classification scheme is based on how arrival and service rates vary over time.
For instance, the signalized intersection example previously presented had the characteristics of a
constant arrival rate over time and a varying service rate over time.

A proposed classification scheme is shown in the figure below, which is a 2 x 2 matrix resulting
in four cells. Each cell can then further be subdivided into subclasses of queueing patterns.
Consider the upper left-hand cell that represents the pattern of a constant arrival rate and a constant
service rate over time. Two rather less interesting sub patterns are encountered. If the arrival rate
is less than the service rate, no queuing is ever encountered. If on the other hand, the arrival rate
is greater than service rate, the queue has a never-ending growth with queue length equal to product
of time and the difference between arrival and service rates.

The upper right-hand cell represents the pattern where the arrival rate is constant over time while
the service rate varies over time. It should be noted that the service rate must be less than the arrival
rate for some period of time, but greater than the arrival rate for other periods of time. Again, this
cell can be further subdivided into subclasses of queueing patterns. Two are shown in the figure
above, but the service rate does not have to be in the form of square-wave. That is, several changes
in service rates of different amounts can be encountered, and the changes do not have to occur
instantaneously but during transitional periods. The signalized intersection example fits the
diagram to the left in this cell, while the occurrence of an incident or accident would result in a
diagram similar to that shown to the right in this cell.

The lower left-cell represents the pattern where the arrival rate varies over time, while the service
rate is constant over time. For queueing to occur and then be dissipated, the arrival rate must be
greater than the service rate for some periods of time and less than service rate during other periods
of time. Two subclasses of queueing patterns have been shown in the figure above. The one on the
left in the cell indicates a square-wave type of arrival rate, while the one on the right in the cell
provides for transitional periods during changes in arrival rates.

The lower right-hand cell represents the more complex situation where both arrival and service
rates vary over time. For queueing to occur then be dissipated, the arrival rate must exceed the
service rate and the later be less than service rate. Two subclasses of queueing patterns are shown
in the figure above. The one on the left cell indicates a square- wave type of arrival and an inverted
square-wave type of service rate. The diagram on the right-hand side is an extension of the first
one with transitional periods during changes in arrival and service rates. Analysis of these more
involved queueing patterns are extensions of queueing patterns that are described in the next two
sections. Simulation is often employed when these more complex queueing patterns are
encountered particularly when sensitivity of parameter values is to be investigated.

Varying service rate problems

The figure above and the accompanying text suggested two possible queueing patterns when
service rate vary over time while arrival rates are constant. Signalized intersections and at-grade
highway-railroad crossings are examples of one queueing pattern, while an incident or an accident
site could cause other queueing pattern. The essential difference between these two cases is
whether the reduced service rate goes to zero or not. Another less obvious difference is whether
the service rate variation is repetitive, such as at a fixed-time traffic signal, or whether it is not
repetitive, such as at a highway-railroad at-grade crossing, incident or accident site, or at a traffic-
responsive controlled traffic signal. Since the signalized intersection was addressed, attention is
now given to queueing analysis at the site of an incident.

The queueing diagram for an incident situation is shown in the figure below;
The figure provides for all input requirements needed to solve this problem. The arrival rate (𝜆 )
is specified in vehicles per hour and is constant for the study period. The normal service rate
(without an incident) is indicated in the diagram as μ, and since it exceeds arrival rate, no queuing
would normally exist. However, an incident occurs that reduces the service rate to μR which is
below the arrival rate, and this lower service rate maintained for t R hours. As in the case for most
highway situations, a “first in, first out” (FIFO) queue discipline is assumed.

In the figure below, a cumulative vehicle versus time diagram is constructed. The arrival is shown
as a straight line passing through the origin with a slope up and to the right equivalent to arrival
rate (𝜆). For the first period of time the service rate becomes equivalent to μR and maintains a flatter
slope until the incident is removed. Then the service rate increases to μ, and the service line has a
steeper slope. This continues until the arrival line and service line intercept, at which the service
line once again overlays the arrival line.
5.4 STOCHASTIC QUEUEING MODELS AND APPROACHES TO DELAY PROBLEMS

5.4.1 INTRODOCTION

So far, the queueing models presented here have assumed that arrival (and possibly) service
functions vary over time in some known fashion. Such queueing models are known as
deterministic queueing models. They are appropriate for many transportation applications because
of the prevalence of distinct traffic peaks in which the capacities of servers are exceeded on a
regular basis. Much of the literature of queuing theory, however, concentrates on a different
situation, in which there are constant long-term average arrival and service rates, but there are also
short-term random fluctuations around these average rates. Under these conditions, arrival rates
will exceed service rates for short time intervals, and queue will form. Models of this type of
queueing are referred to as stochastic queueing models.

Stochastic queueing models are classified as to assumptions made about the distributions of the
arrivals and service times and the number of channels (parallel servers) involved. For instance, an
M/D/1 queueing system is one in which arrivals are exponentially distributed (that is, the number
of arrivals during a given time interval follows a Poisson distribution; the negative exponential
distribution is the distribution of times between arrivals under this assumption), the service rate is
deterministic (that is, no random variation), and there is one channel. Features of interest for
stochastic queuing systems include;

 The average queue length (Q),

 The average waiting time (w),
 Average total delay time spent in system (queue waiting time + service time),(t),
For M/D/1 these are given by;

2𝜌 − 𝜌 2
𝑄=
2(1 − 𝜌)

𝜌
𝑤=
2𝜇(1 − 𝜌)

2−𝜌
𝑡=
2𝜇(1 − 𝜌)

Whereby; 𝜆 = arrival rate

 𝜇 = service rate

𝜌 = 𝜆/𝜇 = traffic intensity.

The most commonly encountered stochastic queuing model is the M/M/1 model, in which both
arrivals and service times are exponentially distributed and there is one channel. For the M/M/1
model, the average queue length, waiting time and total delay are given by;

𝜌2
𝑄=
1−𝜌

𝜆
𝑤=
𝜇(𝜇 − 𝜆)

1
𝑡=
𝜇−𝜆

Stochastic queuing systems are said to be under-saturated when traffic intensity is less than 1.0
and to become saturated when it reaches or exceeds 1.0. Note that for both of the models presented
above; queue lengths, waiting times, and total delays approach infinity as the traffic intensity
approaches 1.0. This is as a result of the assumptions that the arrival rate continues for an indefinite
period of time. Because of this feature, stochastic queueing models are not appropriate for
situations in which there is a definite nonrandom peak in which the arrival rate exceeds the capacity
of the server. On the other hand, deterministic queuing models ignore the random fluctuations in
arrivals and services times which are almost always present and so underestimate queue lengths
and delays. Neither, model is accurate representation of conditions during the transition from
under-saturated to saturated conditions. For this situation model known as the diffusion
approximation has been proposed.

5.4.2 DERIVATION

Output variables

 Utilization rate 𝜌( server utilization, percentage of time that server is busy, where c = the
number of parallel servers).
 Probability of n customers in the system Pn.
  Average number of customers in the system L (service and queue).
 Average number of customers in the queue Lq.
 Average time spent by a customer in the system W (service and queue).
 Average time spent by a customer in the queue Wq.
Transient versus steady state behavior

 Transient behavior (from t = 0).

Performance indicators such as average waiting time, average number of customers in the
queue, etc. are dependent on time, e.g. Wq(t), Lq(t).
 Steady state (stationary) behavior (t → ∞).
Performance indicators such as average waiting time are not dependent of the time
anymore; the probability that the system is in a certain state is completely independent of
time, e.g. Wq, Lq.
Transient state case:
 History of number of customers in the system = graph of number of customers versus time.
 Important to know the queueing strategy disciplines;
- FIFO (Fist in First Out),
- LIFO (Last in First Out = stack),
- SIRO (Service in Random Order),
- SPT (Shortest Processing Time First),
- PR (Priority).
Markov models - for Steady state case:

 Arrival times are exponentially distributed (Poisson process).

 Formulas for steady-state situation.
 Service time are drawn from deterministic, exponentially or Earlang distribution.
 M/M1, M/G/1, M/E K/1, M/D/1.
 M/M/1/N, M/M/c, M/M/1/K/K.
 Formulas for e.g. Pn, L, Lq, W, Wq are derived as follows;
 Formulas for M/M/1 system can be derived quite easily,
𝝀
 Utilization factor of M/.../1, (one channel) 𝝆=𝝁
 𝝀
 Utilization factor of M/.../c, (c channels) 𝝆 = 𝒄𝝁

 If 𝜌 > 1 then the system is unstable; on average more customers arrive than the system can
handle.
 A ‘traffic intensity’ a is used for systems with finite population.
Little’s Equation

 System in stationary condition,

 Number of customers in system L,
 Average total time in system W,
 Little’s Equation; 𝐿 = 𝜆.W
Little’s Law
 Little’s Law states:
Mean number tasks in system = mean arrival rate x mean response time
o As observed before, Little was first to prove this law/relationship
o Applies to any system in equilibrium, as long as nothing in black box is creating
or destroying tasks
o For a Poisson process with average arrival rate, the probability of seeing n arrivals
in time interval delta t (δt)

Markov model: M/M/1

 Average number of customers in the system for M/M/1 in steady state condition

 Steady state, so P0 *1 = P1 *m
 But also; P0 *1 +P2 *m = P1 * (1+m)
 Pn = 1 n/m * P0
 For the sum of P, the following holds;
 𝑛=∞

∑ 𝑃𝑛 = 1
𝑛=0

 The relation between Pn and P0 is;

𝑛=∞ 𝑛=∞
𝜆𝑛 𝜆𝑛
∑ ∗ 𝑃0 = 1 = 𝑃0 ∑
𝜇 𝜇
𝑛=𝑜 𝑛=0

𝜆𝑛
 If the system in the steady state then 𝜆/ 𝜇 ˂ 1 and: ∑𝑛=∞
𝑛=𝑜 𝜇 can be replaced by;

1/ (1- 𝜆/ 𝜇)
𝜆
 Using this, we can calculate tata P0 equals; 𝑃0 = 1 − 𝜇

𝜆 𝜆𝑛
 Thus, we get for Pn; 𝑃𝑛 = (1 − 𝜇)(𝜇 )

 Because for L we know that;

𝑛=∞ 𝑛=∞
𝜆 𝜆𝑛 𝜆 𝜆 𝜆𝑛
𝐿 = ∑ 𝑛 ∗ 𝑃𝑛 = ∑ 𝑛 ∗ (1 − ) ∗ ( ) = ∗ (1 + + +⋯)
𝜇 𝜇 𝜇 𝜇 𝜇
𝑛=0 𝑛=0

 Writing it out yields;

𝜆
𝐿=
𝜇−𝜆

 Fill in and use Little’s equation. Then we get the table with the most important values for
M/M/1 system.
𝜆
𝐿=
𝜇−𝜆
𝐿 1
𝑊= =
𝜆 𝜇−𝜆
1 𝜆
𝑊𝑄 = 𝑊 − =
𝜇 𝜇(𝜇 − 𝜆)
𝜆2
𝐿𝑄 = 𝜆𝑊𝑄 =
𝜇(𝜇 − 𝜆)
𝜆 𝜆𝑛
𝑃𝑛 = (1 − )( )
𝜇 𝜇
Exercise

On average one customer per minute arrives (number of customers per minute Poisson distributed).
Average service time is 40 seconds per customer (exponentially distributed). What is;

- Average time in system?

- Average waiting time?
- Average number of customers in the queue?
- Probability there are exactly 5 customers in the system?

M/D/1 system

 The arrival times are exponentially distributed,

 Service times have no variance,
Steady state parameters for M/D/1 can be calculated from M/G/1 by substituting 𝜌2 = 0.
 Note: the average queue length for M/D/1 is exactly half M/M1
1 𝜌2
𝐿=𝜌+ ( )
2 1−𝜌
𝐿 −1
1 𝜌𝜇 −1
𝑊= =𝜇 + ( )
𝜆 2 1−𝜌
1 1 𝜌𝜇 −1
𝑊𝑄 = 𝑊 − =
𝜇 21 − 𝜌
1 𝜌2
𝐿𝑄 = 𝜆𝑊𝑄 =
21−𝜌
Exercise
Planes that land on a runway, one runway,30 arrivals per hour (Poisson distributed), 90
seconds deterministic landing time. Fuel cost f. 5000- per hour. Calculate;
- Average length of queue,
- Average waiting time,
- Expected total number in the system,
- Fuel cost per hour as a result of delay in queue.
Reducing waiting times

 Reduce number of arrivals per unit time,

  Reduce service time,
 Increase number of servers,
 Decrease variance in service times.
M/M/1/N system

 If the system is full the customer leaves,

 In this case 𝜆/ 𝜇 is called a,
 𝜆e is the arrival intensities of the entities that stay in the system.
 𝜆e ˂ 1
 Calculation: 𝜆e = 𝜆( 1-PN)
L {a[1-(N+1) aN+NaN+1] 𝜆≠ 𝜇

N/2 𝜆= 𝜇

1-PN {1-aN/1-An+1 𝜆≠ 𝜇

{N/N+1 𝜆= 𝜇

𝝀e 𝜆(1-PN) = 𝜇(1-P0) = 𝜇 e

𝝆 𝜆e/ 𝜇= 1-P0

W L/𝜆e

WQ W-1/ 𝜇

LQ 𝜆eWQ = L-(1-P0)

Pn {(1-a) an/1-aN+1 𝜆≠ 𝜇

𝑛 = 0,1, … , 𝑁

{1/…… 𝜆= 𝜇

Exercise
Small travel agent with one employee, 3 chairs to wait, 𝜆= 5 customers per hour (Poisson), 𝜇= 10
customers per hour (Poisson). Calculate;

- The effective arrival intensity 𝜆e.

M/M/c system

 More servers,
 If the number of customers in the system n ˂ c then new arrival can be served immediately,
 Stable system if 𝜆˂ c𝜇,
 Utilization rate is not 𝜌 = 𝜆 /𝜇 but 𝜌 = 𝜆/c 𝜇,
 If 𝜌 > then the system grows with ( 𝜆 - c 𝜇),
 Complex formulas, so often tables or graphs are used.

Exercise
Customers at a post office. 1 queue and more servers. Arrival intensity 2 per minute (Poisson),
service time 40 seconds (exponential). Determine;

- Number of servers needed to reach steady state,

- Probability that there are no customers in the system with the calculated number of
servers,
- Average queue length with the calculated number of servers.
5.4.3 STOCHATIC QUEUEING ANALYSIS

A flow chart depicting various queuing situations is shown in figure below. If the arrival
distribution and /or service distribution is probabilistic, the exact arrival and /or service time of
each vehicle is unknown and stochastic queuing analysis rather than deterministic queueing
analysis must be selected.

To use stochastic queuing analysis, the traffic intensity (𝜌) must be less than 1. Traffic intensity
is;

𝜆
𝜌=
𝜇

Where 𝜌 = traffic intensity

𝜆 = mean arrival rate (vehicles per time interval)

 𝜇 = mean service rate per channel (vehicle per time
interval)

The arrival and service rates can be expressed in a variety of ways. For example, the time interval
maybe specified in second, hours and so on, but the arrival and service rates must employ the same
time units. In some cases, mean arrival times and /or, mean service times are given and require
conversion to arrival and service rates as shown below;

3600
𝜆=
ℎ̅

3600
𝜇=
𝑠̅

Where ℎ̅ = mean arrival time (seconds per vehicle)

𝑠̅ = mean service time (seconds per vehicle)

3600 = a constant to convert arrival and service rates to

hourly rates

If the traffic intensity is greater than 1, the only possible solution approaches are the convert the
queuing process to a deterministic queuing problem or to introduce multi-time slice, varying mean
arrival rates, and mean service rates and solve using microscopic simulation techniques.
Deterministic queuing analysis was discussed and simulation techniques will be discussed alter.

Within stochastic queuing analysis there are two approaches;

i) If the analyst wishes to study stochastic queuing process macroscopically, an analytical

approach based on sets of mathematical equations can be utilized.
ii) On the other hand, if the analyst wishes to study the stochastic queueing process
microscopically, microscopic simulation techniques are required.

The reminder of this chapter is devoted to analytical approach of solving stochastic queuing
problems.
There are many types of probability distribution of arrival and service rates. A classification
scheme based on a few of more common distributions is shown in the Table below. If the arrival
and service distributions both have constant mean values, a deterministic queuing approach is
employed. The letter D denotes a constant mean value, while M, E and G represent random,
Earlang, and generalized forms of probability distributions.an unlimited number of probability
distributions can be utilized, but sets of mathematical equations have been derived for only a few
of the more common probability distributions.

Referring to the table, each cell is denoted by two letters (i.e. M/D). This represents a stochastic
queuing distribution situation in which arrivals are randomly distributed while service rimes are
constant. This code identification scheme can be further extended by identifying the number of
service channels, queue limits, and the queuing discipline characteristics. For example;

M/D/1 (∞, FIFO)

identifies the same stochastic queuing problem described above and in addition indicates there is
a single service channel with queue discipline characteristics of permitted queue length of infinity
with no diversion and a “first in, first out” service system. If only a code of M/D is given, it is
understood to represent M/D/1 (∞, FIFO).

Three stochastic queueing situations based on the analytical approach are presented in the next
three sections. The first two deal with single-channel services, with random arrival distributions
and with random or constant service distributions [M/M/1(∞, FIFO) and M/D/1 (∞, FIFO)]. The
final section is devoted to Multi-channel systems [M/M/N (∞, FIFO)].
Random service problem

The set of equations used for stochastic queuing problem in which both the arrival and service
times are randomly distributed is shown in the table below. The discussion in this section is
devoted to single-service channel with no limit on queue length and “first in, first out” service.

Queuing characteristics M/M/1 (∞, Alternative expression

Symbol Definition FIFO)
Equation
𝜌 Traffic intensity 𝜆/μ 𝜆
𝐼 = 𝑃0 = 1 −
𝜇
P (0) Probability of empty system 1- 𝜌 -
P(n) Probability of exactly n units 𝜌n(1-𝜌) -
in the system
E(m) Average number waiting to be 𝜌2/1- 𝜌 𝜆2 𝜌2
𝐿𝑞 = =
served 𝜇(𝜇 − 𝜆) 1 − 𝜌

E (m/m > 0) Average number waiting when 1/1- 𝜌 -

queue is present
E(n) Average number in system 𝜌/1- 𝜌 𝜆
𝐿=
(waiting + service) 𝜇−𝜆

Var (n) Variance of E(n) 𝜌/ (1- 𝜌)2 -

E(v) Average time in the system 1/μ (1- 𝜌) 1
𝑤=
𝜇−𝜆
E(w) Average waiting time only 𝜌/μ (1- 𝜌) 𝜆
𝑤𝑞 =
𝜇(𝜇 − 𝜆)

The traffic intensity, 𝜌, is the ratio if mean arrival rate to the mean service rate and can vary from
zero to 1. As 𝜌 increases the various queuing performance measures increase at an increasing rate
until 𝜌 = 1, the various measures go to infinity. The probability of some or more vehicles being in
the system at any point in time is equal to traffic intensity. Therefore, the probability of an empty
system is equal to (1- 𝜌). If the analyst is untested in more specific information about the number
of vehicles in the system, Gerlough and Huber have shown that;
 𝑃 (𝑛) = 𝜌 𝑛 (1 − 𝜌)

Where P(n) = probability of exactly n vehicles in the system

𝜌 = traffic intensity

n = number of vehicle sin system

The probabilities of n or fewer vehicles in the system as a function of traffic intensity are shown
the figure below. For example, if the arrival rate is 100 vehicles per hour and service rate is 150
vehicles per hour, the traffic intensity is equal to 0.67. From the figure, the probability of exactly
n vehicles in the system and cumulative probabilities can be obtained. For example, the probability
of exactly 1,1,2,3,4 and 5 vehicles in the system is 0.33, 0.22, 0.15, 0.10, 0.07 and 0.04
respectively. Note that as 𝜌 approaches 1, the probabilities of a large number of vehicles in the
system increases very rapidly.

The queue length characteristics as a faction if traffic intensity is shown the follow figure. Using
the previously calculated traffic intensity value of 0.67 and the equations shown in the previous
table, the queue length characteristics would be;

E(m) = 1.4 vehicles

E(n) = 2.0 vehicles

E (m/m > 0) = 3.0 vehicles

Var (n) = 6.2 vehicles

Without careful though one might assume that the difference between the average number in the
system [E(n)] and the average number waiting to be served [E(m)] should be equal to one of single-
channel, server. However, there is a basis because of the probability of an empty system. Infact, it
can be shown that;

𝐸 (𝑛) = 𝜌 + 𝐸(𝑚)

Thus, the difference varies from zero (when 𝜌 = 0) to 1(when 𝜌 = 1).

Special consideration should also be given to the relationship between the average number waiting
when a queue is present [E (m/m > 0)] and the average number waiting to be served [E(m)].
Contrary to initial intuition, the greatest difference occurs under low traffic intensity, and the
difference approaches zero when traffic intensity approaches 1 because the probability of an empty
system approaches zero.

In reference to the average number in the system [E(n)] and its variance [Var (n)], they are
approximately equal under low traffic intensity, while the ratios becomes 1 to 10 when the traffic
intensity equals 0.9.

The effects of arrival and service rates on average waiting time and total waiting time are shown
figures below. The two figures are constructed in a similar manner, when the mean arrival is plotted
on the vertical axis and the mean service rate on the horizontal scale. Two sets of contour lines are
constructed on the figures. The traffic intensity continue line appear as radial lines extending up
and to the right from the origin. Steeper slopes represent highest traffic intensities. The second set
of contours lines represent average waiting time in first figure and the total system waiting time in
the second figure. The two diagrams clearly show how waiting times increase rapidly as traffic
intensity approached unity.

Constant service problem

The set of equations for a stochastic queuing analysis problem in which arrivals are randomly
distributed and the service rates is constant is known shown in the table below. The discission in
this section is devoted to single-channel service with no limit on queue length and ‘first in, first
out” service.

Queuing characteristics M/D/1 (∞, FIFO) Alternative

Symbol Definition Equation expression

𝜌 Traffic intensity 𝜆/μ -

P (0) Probability of empty - 𝑃0 = 1 − 𝜌
system
P(n) Probability of exactly - -
n units in the system
E(m) Average number - 𝜌2
𝐿𝑞 =
waiting to be served 2(1 − 𝜌)
 E (m/m > 0) Average number - -
waiting when queue is
present
E(n) Average number in 2𝜌 – 𝜌2/2(1- 𝜌) 𝐿 = 𝐿𝑞 − 𝜌
system (waiting +
service)
Var (n) Variance of E(n) - -
E(v) Average time in the 2- 𝜌/2μ (1- 𝜌) 1
𝑊 = 𝑊𝑞 +
system 𝜇

E(w) Average waiting time 𝜌/2μ (1- 𝜌) 𝐿𝑞

𝑊𝑞 =
only 𝜆

As an example, the queuing performance in single-channel system with random arrivals and
constant service. The figure below demonstrates the effects of arrival and service rates on average
number and average time in the system. The contour lines in first figure represent the average
number of vehicles in the system and extend up to the right from the origin. Traffic intensity (𝜌)
can also be shown as contour lines in the same diagram. The diagram clearly shows the rapid
increase in the average number in the system as traffic intensity approaches a value of 1.

The second figure shows contour lines representing average vehicle time in the system. This
relationship is a little more complex since the average time is a function of the traffic intensity and
mean service rate. The diagram clearly shows the rapid increase in the average time in the system
as traffic intensity approaches a value of 1 and as mean service time approaches zero.
Examples

M/D/1
Question

Customers arrive at a ticket counter of a local movies theater at the rate of 165 persons per hour
over the period until, the movie starts, while the service rate is 195 persons per hour. Compute the
characteristics of the system.

Solution
165
- Mean arrival rate: 𝜆= = 2.75 𝑝𝑒𝑟𝑠𝑜𝑛𝑠 /𝑚𝑖𝑛
6

(exponential)
- Constant service rate: 𝜇 = 3.25 𝑝𝑒𝑟𝑜𝑠𝑛𝑠 /𝑚𝑖𝑛
𝜌 = 0.846

𝜌2 0.8462
- Mean number of units in the system: 𝐿𝑞 = 2(1−𝜌) = 2(1.154) = 2.327 𝑝𝑒𝑟𝑠𝑜𝑛𝑠

- Mean number of units in the system: 𝐿 = 𝐿𝑞 + 𝜌 = 3.173 𝑝𝑒𝑟𝑠𝑜𝑛𝑠

𝐿𝑞 2.327
- Mean number in queue: 𝑤𝑞 = = = 0.846 𝑚𝑖𝑛
𝜆 2.75
1
- Mean number in the system: 𝑊 = 𝑤𝑞 + 𝜇 = 0.846 + 0.308 =

1.154 𝑚𝑖𝑛
- Service facility utilization factor: 𝜌 = 0.846 𝑚𝑖𝑛
 - Probability of no units in the system: 𝑃0 = 1 − 𝜌 = 0.154 𝑚𝑖𝑛
M/M/1

Question

A movie theatre ticket both has a mean arrival rate 3 person/minute and service rate is 4 persons
/minute. Calculate the characteristics of this queuing system applying the M/M/1 model.

Solution

𝜆
𝜆 = 3; μ = 4; 𝜌 = 𝜇 = 0.75

𝜆 3
𝐿 = 𝜇−𝜆 = 4−3 = 3 𝑝𝑒𝑟𝑠𝑜𝑛𝑠

𝜌2 0.752
𝐿𝑞 = 1−𝜌 = 1−0.75 = 2.25 𝑝𝑒𝑟𝑠𝑜𝑛𝑠

1 1
𝑊 = 𝜇−𝜆 = 1 = 1.00 𝑚𝑖𝑛

𝜆 3
𝑊𝑞 = 𝜇(𝜇−𝜆) = 4(1) = 0.75 𝑚𝑖𝑛

𝜆
𝐼 = 1 − 𝜇 = 1 − 0.75 = 0.25

Economics of operating characteristics of Queuing discipline

Engineers and managers making decisions regarding the level of service that needs to be
maintained in handling goods, vehicles and people often have o consider the overall savings that
would accrue if an additional worker were to be put in operation. After all, the savings of
customer’s tike is highly important for a business that intends to increase its clientele. The cost
trade-off relationship between the total cost of the facility and the level -of- service is shown in
the figure below. The objective o to minimize the total cost of service and waiting time to achieve
respectable service level.
Multichannel problem

In the previous discussion a single file of arriving vehicles being served a single channel was
addressed. If the multichannel server is provided (such as several tolls at a toll plaza), two types
of entry control could be implemented. One type would be to introduce as many entry channels as
service channels and divide the arrival demand equally between channels. That is, the arrival rate
of each channel would be equal to the total arrival rate divided by the number of channels. The
service rate for each channel would be equal to the total service rate divided by the number of
channels. The traffic intensity for each individual channel or for the combination of all channels
would be;

𝜆⁄𝑁 𝜆
𝜌= =
𝜇 ⁄𝑁 𝜇

Where N = is the number of channels.

The key feature of this approach is that once the arriving traffic divides equally between channels
serving the various service areas, the traffic cannot change channels even if their channel is queued
and a parallel channel is empty. This causes inefficient use of the service area and results in greeter
delays and loner queues.

Another type of entry control can overcome this inefficient use of service area by combining the
arrival traffic into one channel and then selecting the lead vehicle which is waiting, to go to the
available empty service channel. This will keep the service are busy as long as the queue is present.
The difficulty is the practical implementation of such an entry control scheme. Probably the most
successful examples of such an entry control schemes are at banks and in airport check-in counters.

An example problem is presented to demonstrate the application of the two entry control schemes
and to provide a comparison of their performance. In this example, assume a total arrival rate of
800 vehicles per hour with the arrival randomly distributed. Two toll gates are available for service,
and each toll gate a service time of 6 seconds, which is also randomly distributed. There is no limit
on the length of the queue. The difference is the entry control scheme, that is, how the vehicles are
served.

- Consider the first entry control scheme in which the arriving vehicles are distributed
equally between two toll gates. In concept, there is a barrier between the two toll gates,
and once a vehicle is assigned a toll gate service channel it cannot change to the other
channel. Therefore, each channel is independent of the other, and the arrival rate and
service tomes per channel are calculated and analyzed as a single-channel problem as
discussed earlier. In essence this multichannel problem is converted to a single- channel
problem, M/M/1(∞, FIFO). The arrival rate per channel is 400 vehicles per hour and
the service rate per toll gate is 600 vehicles per hour. The traffic intensity (𝜌) is
calculated as 0.67. the various queueing performance characteristics can be calculated
using previously discussed equations. These results and a comparison of results with
the other entry control scheme are discussed next.
- Now considering the other entry control scheme, where a single arrival queue is
formed and the lead vehicle in the queue is always selected to go to the first available
service channel. This is a multichannel problem that is analyzed as a multichannel
problem and classified as M/M/2(∞, FIFO). The total arrival rate (𝜆) is 800 vehicles
per hour, the service(μ) per channel is 600 vehicles per hour, and the total service
rate(μN) is 1,200 vehicles per hour. Note that while 𝜆 denotes the total arrival rate, μ
denotes the service rate per channel and μN denotes the total service rate. The queueing
performance equations for random arrival-random service distribution system (where
N > 1) are summarized in the table below.
 Queuing characteristics M/M/N (∞, FIFO)
Symbol Definition EQUATION
𝜌 Traffic intensity 𝜆
𝜇
Uf Utilization factor 𝜌
𝑁
P(0) Probability of an empty system 𝑁−1
𝜌𝑛 𝜌𝑁
∑( )+
𝑛! 1−𝜌
𝑛=0 𝑁! [ 𝑁 ]

P(n) Probability of exactly n units in the 𝑃 (0) 𝜌 𝑛 𝑃0

𝑜𝑟 ( 𝑛−𝑁 )
system (n ≤ N or n ≥ N) 𝑛! 𝑁 𝑁!

E(m) Average number waiting to be served 𝑃(0)𝜌 𝑁+1 1

(
𝑁! 𝑁 [(1 − 𝜌)/𝑁]2
E (m/m Average number waiting when queue 1
/𝑁
> 0) is present (1 − 𝜌)

E(n) Average number in system (number 𝜌 + 𝐸(𝑚)

waiting + number in service)
E(v) Average time in system 𝐸(𝑛)
= 𝜋𝑟 2
𝜌𝜇
E(w) Average waiting time only 1
𝐸 (𝑣 ) −
𝜇

The traffic intensity (𝜌) can be calculated to be 1.33. a new term utilization factor (Uf) is
introduced, which is the ratio of the total arrival ret divided by total service rate, which in this
example is equally to 0.67. While 𝜌 can exceed 1.0, Uf cannot.

The queueing performance results of the two entry control schemes are summarized in the table
below.

Queuing characteristics M/M/2 (∞, FIFO) Result

Symbol Symbol Independent Shared
channels channels
𝜌 Traffic intensity 0.67 1.33
 Uf Utilization factor 0.67 0.67
P (0) Probability of an empty system 0.33 0.20
P(n) Probability of exactly 1, 2 and 3units 0.22 0.27
in the system 0.15 0.18
0.10 0.12
E(m) Average number waiting to be served 1.36 1.08
E (m/m > 0) Average number waiting when queue 3.00 3.00
is present
E(n) Average number in system (waiting + 2.00 2.41
service)
E(v) Average time in system(seconds) 18.0 10.9
E(w) Average waiting time only(seconds) 12.0 4.9

Example of multichannel problem

An entrance to a tollway has four tollbooths. Vehicles arrive at an average of 1,800 vehicles /hr.
and take 6.67 seconds to pay their tolls. Both arrivals and departures can be assumed to be
exponentially distributed. Compute the average queue length, the time in the system, and
probability of waiting in queue.
Solution
1800
𝜆= = 30 𝑣𝑒ℎ𝑖𝑐𝑙𝑒𝑠/𝑚𝑖𝑛
60
60
𝜇= = 9 𝑣𝑒ℎ𝑖𝑣𝑙𝑒𝑠 /𝑚𝑖𝑛
6.67
30
𝜌= = 3.333
9
3.333
𝜌⁄𝑁 = = 0.8333 < 1
4
Now, if the administration want four booths open, then;
1
𝑃𝑜 = 0.0213
3.333 3.3332 3.3333 3.3334
1+ 1! + + 3! +
2! 4! (0.1667)
0.0213(3.333)5 1
𝐿= [ ] = 3.289 𝑣𝑒ℎ𝑖𝑐𝑙𝑒𝑠
4! 4 (0.1667)2
 3.333 + 3.289
𝑊= = 0.2207 𝑚𝑖𝑛
30
0.0213(3.333)5
𝑃𝑛>𝑁 = = 0.548 𝑚𝑖𝑛
4! 4(0.1667)

5.5 SIMULATION MODELS AND TRAFFIC PROBLEMS

Simulation can be defined as follows:

Different texts interpret simulation differently:

The most basic reasons for using simulation extensively are:

i. The need to test the behavior of a new system or operating procedure prior to its actual
construction (where use of real systems may be expensive or constitute risks)
ii. The need to test alternate systems under identical conditions
iii. One can experiment with various control policies without disrupting real traffic and
making that traffic part of an experiment.
iv. The same can be said of alternative designs; one can experiment with alternative designs
for remedies to existing or future conditions, such lane additions, rerouting of traffic,
changes in direction of links, suitability of junction and intersection designs, etc.
v. Above all, these alternative control implementations and designs can be considered
rapidly, without major capital investment.
 vi. Whenever desired, various alternatives can be considered with exactly the same traffic
load and conditions, something that is simply not feasible in the real world.
Some strengths of simulation are outlined in the Table 5.5.1 below, while simulation
reservations in Table 5.5.2.
Table 5.5.1: Simulation Model Strengths

Table 5.5.2: Simulation Model Reservations

In cases where simulation model itself becomes the end product rather the use of the model, or
where the builder concentrates too much effort in small, less significant aspect of the model and
loses sight of the total model and its use and output, it is not worth the cost and time
Steps in Developing a Simulation Model:
A flow chart of procedural elements of simulation are given in Figure 5.5.1 below.

Figure 5.5.1: Procedural Elements in Simulation

The following give a qualitative guide in development of simulation models for traffic systems.
The ninth element is validation. This element includes verification, calibration, and validation.
The relationship between these three tasks are depicted graphically in Figure 5.5.2. Verification
is a more exhaustive form of the earlier task of debugging code. The objective is to ensure that
the computer code is doing exactly what is specified in the flowcharts, with attention, at time
point, being given to the real-life situation. The next task is calibration, and the difficulty of this
task is dependent on the quality of field data measurements, the comprehensiveness of the
computer model, and the complexity of the system being studied. Only a portion of the collected
input data and output performance cab be used for calibration; the remainder data must be
reserved for validation. Reasonable adjustments are made in the model to match the model’s
output with real-life observations. The final task is validation. The unused portion of the field
collected data input is placed in the model, and the model predictions are compared with the
related unused portion of the field-observed real-life observations. No adjustments are made to
the model and the differences record how well the completed model represents real life under
conditions tested. If the results of the validation are acceptable, then the model is ready for
application. If not, further calibration and validation is required.

Fig. 5.5.2: Verification, Calibration &

Validation Process
Example of simulation scheme for Gap Acceptance:

Figure 5.5.3: Generation of Freeway Gap Acceptance

SUMMARY

The reasons for simulation include (but need not be limited to):
1. The need to test the behavior of a new system or operating procedure prior to its actual
construction:
(a) The construction of the new system may be very expensive and/ or time consuming.
(b) Experimentation with the real system may entail considerable risk (such as traffic accidents).
2. The need to test alternate systems under identical conditions. For instance, it is never possible
to exactly reproduce a specific traffic condition in the field; in simulation it is quite routine to
submit the same traffic conditions to several alternative systems.
3. One can experiment with various control policies without disrupting real traffic and making
that traffic part of an experiment.
4. The same can be said of alternative designs; one can experiment with alternative designs for
remedies to existing or future conditions, such lane additions, rerouting of traffic, changes in
direction of links, suitability of junction and intersection designs, etc.
5. Above all, these alternative control implementations and designs can be considered rapidly,
without major capital investment.
6. Whenever desired, various alternatives can be considered with exactly the same traffic load
and conditions, something that is simply not feasible in the real world.

Any simulation may be divided into the following steps:

1. Formulation of a model.
2. Reduction of the model to a language acceptable by the computer.
3. Program checkout and internal verification of the model.
4. Experiment planning and design:
(a) Design of an experiment that will yield the desired information.1
(b) Determination of how each of the test runs will be executed.
5. Performance of experiment(s): (a) Model validation. (b) Simulation of new conditions.
6. Interpretation of results.

Interpretation of Results
The interpretation of simulation experiment results is similar to the interpretation of any
experimental results. Often the result is the testing of some hypothesis; in other cases, the
interpretation consists of fitting some curve to the output of the experiment.

Conclusion:
In formulating a simulation model, one strives for sufficient realism to adequately describe the
phenomena of interest. However, one should not go "overboard" in including extra details that
will not significantly affect the results. Programs should be properly checked out and validated.
Traffic simulation programs that are properly modeled and validated constitute important
experimental facilities for traffic study.