Queuing and Transportation

Transportation Logistics
Prof. Goodchild
Spring 2009
Two ways to address queues

• Make an analytical model of customers

needing service, and use that model to
predict queue lengths and waiting times
– Steady state assumption
– Simulation
Definitions
• Customers — independent entities that arrive at random times to a Server and wait for
some kind of service, then leave.

• Server — can only service one customer at a time; length of time to provide service
depends on type of service;

• Arrival time: time customer arrives at the back of the queue

• Departure time: time customer leaves server

• Inter-arrival time: time between successive arrivals of customers

• Service time: time for server to serve one customer (amount of time you are delayed if no
one else present)

• Queue — customers that have arrived at server but are waiting for their service to start are
in the queue.

• Queue Length at time t — number of customers in the queue at time t.

Total Time in System

• Service time: the amount of time you would be

delayed if no other customers required service

• Waiting time: the amount of time you have to wait

because others also want service
– The price you pay for others

• Total Time in System = Service time + Waiting

time
Queue Discipline

• FIFO
– Traffic
– intersection
• LIFO
– Elevator
– Airplane
• Random
– Fluids
• Priority
Transportation Applications

• Traffic congestion
• Being serviced at:
– Border
– Toll plaza
– Bus stop
– Goods waiting at a distribution center
– Marine terminal
– ….
Activated
Upstream of bottleneck/server Downstream

Arrivals Departures

Server/bottleneck

Direction of flow
Not Activated

Arrivals Departures

server
Flow Analysis

• Bottleneck active
– Service rate is capacity
– Downstream flow is determined by bottleneck
service rate
– Arrival rate > departure rate
– Queue present
Flow Analysis

• Bottle neck not active

– Arrival rate < departure rate
– No queue present
– Service rate = arrival rate
– Downstream flow equals upstream flow
Queue Analysis – Graphical
Departure
Rate
Delay of nth arriving vehicle
Cumulative Number of Items

Arrival
Rate
Maximum queue

Maximum delay

Total vehicle delay

Queue at time, t1

t1 Time
Queue Notation

Number of
Arrival rate servers

X /Y / N
Departure rate

• Popular notations:
– D/D/1, M/D/1, M/M/1, M/M/N
– D = deterministic
– M = other distribution
Poisson Distribution
• Good for modeling random events
– Standard deviation equals the mean
• Count distribution
– Uses discrete values

P n  
t  n
e  t

n!
P(n) = probability of exactly n vehicles arriving over time t
n = number of vehicles arriving over time t
λ = average arrival rate
t = duration of time over which vehicles are counted
 Example Graph
0.25

0.20
Probability of Occurance

0.15

0.10

0.05

0.00
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Arrivals in 15 minutes
 Example Graph
0.25

Mean = 0.2 vehicles/minute

0.20
Probability of Occurance

Mean = 0.5 vehicles/minute

0.15

0.10

0.05

0.00
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Arrivals in 15 minutes
• If we assume Poisson arrival process
• Inter-arrival times are exponentially
distributed
 Example: Arrival Intervals
1.0

0.9 Mean = 0.2 vehicles/minute

0.8 Mean = 0.5 vehicles/minute
Probability of Excedance

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0
0 2 4 6 8 10 12 14 16 18 20
Time Between Arrivals (minutes)
Little’s Formula (1961)

E ()  E (  )
• T = time spent by a customer in the queueing system
•  = arrival rate
• N = number of customers in the system

• The long-term average number of customers in a stable

system N, is equal to the long-term average arrival rate, λ,
multiplied by the long-term average time a customer
spends in the system, T
• Steady state assumption
Steady State Analysis

   1.0


• M/D/1
2
– Average length of queue Q
21   

1   
– Average time waiting in queue w   
2  1   

1 2  
– Average time spent in system t  
2  1   

λ = arrival rate μ = departure rate =traffic intensity

Queue Analysis

   1.0


• M/M/1
2
– Average length of queue Q
1   

1  
– Average time waiting in queue w   
    

1
– Average time spent in system t
 

λ = arrival rate μ = departure rate =traffic intensity

Queue Analysis

• D/D/1
– Average length of queue

– Average time waiting in queue

– Average time spent in the system

Queue Analysis

  N  1.0


• M/M/N
P0  N 1  1 
– Average length of queue Q  
N! N  1   N 2 

 Q 1
– Average time waiting in queue w  
 

 Q
– Average time spent in system t


λ = arrival rate μ = departure rate =traffic intensity

M/M/N

  N  1.0
– Probability of having no vehicles 
1
P0  N 1 nc
 N

nc  0 nc !

N !1   N 

– Probability of having n vehicles

 P0
n  n P0
Pn  for n  N Pn  n N
for n  N
n! N N!
– Probability of being in a queue
P0  N 1
Pn N 
N! N 1   N 
λ = arrival rate μ = departure rate =traffic intensity
 Queue times depend on variability

items

time
Can’t store extra capacity

• No reservoir for storing capacity

• If capacity goes unused, it is wasted
Queue times depend on variability

Delay will be very different depending on the arrival PATTERN, not just
number of arrivals
limitations

• There are many cases when we want to

consider changes to the arrival rate
• This is difficult to do when you are limited
to steady state assumptions
• Limited number of distributions that
provide a closed form expression
Simulation

• In general we are interested in the

variability in arrival rates or service times
• If these are constantly varying a steady
state assumption is fine
• The alternative is to use a discrete event
simulation framework and keep track of
individual customers
– Microsimulation of an intersection
– Queue simulation
Queue simulation

• Simulation based approach

• Track vehicles
• Step through time
• Can change arrival rates, service times,
with knowledge of previous system state
• Border wizard
Examples

• Marine terminal
• Rail infrastructure
• International border
• Airport terminal
Port gate and terminal stacks

Stacks

Gate
Observed data
Theoretical wait times

Single server, deterministic service times

0.1

0.09

0.08

0.07

0.06
Wait time

0.05

0.04

0.03

0.02

0.01

0
10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Arrival rate
 Rail line as server
Average Annual Delay (container hours per through container)

18.00
16.00
Container hours delay per

14.00
container through

12.00
10.00 7 days a week operation
8.00 5 days a week operation
6.00
4.00
2.00
-
1,000,000 2,000,000 4,000,000
TEU throughput

Bottleneck activation
Airport of the Future

• Separates queue into

two different processes
– Check in
– Bag check

• Allows travelers to
enter mid-stream
Change in terminal processing

Baggage flow behind the counter

B B B B B B
K K K K K K K K K K K K

Baggage flow
Queue approaching the counter

B B

P
B B

B B

B B

K K K K K K K K

P
Previous system

K B

K B

K B
Airport of the future

K B

K B

K B
Service times got worse!

Total Service Time 2008 2005

Mean 07:21 04:59
Median 06:35 04:09
Standard Deviation 04:46 03:19
Sample Variance 00:01 00:00
Kurtosis 18:16 07:09
Skewness 12:04 58:31
Range 31:03 19:45
Minimum 00:01 00:38
Maximum 31:04 20:23
Count 510 352
Confidence Level(95.0%)-max 07:46 05:20
Confidence Level(95.0%)-min 06:56 04:38

Does not include wait time, only measured from arrival at check-in desk
Total times do improve with AF

• Encourages travelers to check-in online

• Reduces perceived wait time
– Start process sooner
– Can’t see a big queue
• Reduces employee requirements
• Improves space utilization
Border as server
Observations from one day
Regression equations by DOW
Regression equations by Season
Number of very long delays
Proportion of very long delays
Transportation Realities

• For many systems other factors influence

delay
• Queing can be used to model wait times
• Appropriate tool can be identified on a
case by case basis
• Be sure you understand the theoretical
and practical framework