Course Outline

 Linear Programming problems – Formulation, Graphical

method - Sensitivity Analysis and Duality
 Transportation, Transshipment and Assignment
problems
 Decision trees
 Network optimization models
 Integer programming problems
 Project management
 Simulation – A brief overview
 Waiting line models
 Multi-objective decision problems
 Dynamic Programming
Course Outline
 Linear Programming problems – Formulation, Graphical
method - Sensitivity Analysis and Duality
 Transportation, Transshipment and Assignment
problems
 Decision trees
 Network optimization models
 Integer programming problems
 Project management
 Simulation – A brief overview
 Waiting line models
 Multi-objective decision problems
 Dynamic Programming
Integer Programming
Problems
 Knapsack Problem
 Fixed charge Transportation Problem
 Facility Location Problem
 Set-Covering Problem
 Traveling Salesman Problem (TSP)
 SUDOKU
 Convoy Movement Problem (CMP)
How are IP problems
different?
Why are IP problems
complex?
Sensitivity Analysis for IPP
 Integer programming problems do not
readily lend themselves to sensitivity
analysis as only a relatively few of the
infinite solution possibilities in a feasible
solution space will meet integer
requirements
Some interesting info on
IP…
 Turkish Petroleum Refineries Corporation used BIP to
analyze capital investments worth tens of millions of dollars
to expand refinery capacity – Interfaces 2020
 South African National Defense Force – OR study to upgrade
capabilities – Interfaces 2017
 United States Air Force Space Command – Interfaces 2003
 Grantham, Mayo, Van Otterloo and Company – Portfolio
management representing $8 billion in assets – Interfaces
1999
Knapsack Problem
 A thief breaks into a museum. Fabulous paintings,
sculptures, and jewels are everywhere.
 The thief has a good eye for the value of these objects,
and knows that each will fetch hundreds or thousands of
dollars on the clandestine art collector’s market.
 But, the thief has only brought a single knapsack to the
scene of the robbery, and can take away only what he
can carry.
 What items should the thief take to maximize the haul?
Knapsack Problem
 More formally:
 The thief must choose among n items, where the i th

item is worth Vi dollars and weighs Wi pounds

 Carrying at most ‘K’ pounds, maximize the value of
knapsack
 Note: assume Vi, Wi, and K are all integers

 An item must be taken or left in entirety

 Applications: Resource utilization, Capital investments

and Financial portfolios, Cryptography etc
Knapsack Problem – Mathematical
Model
n
Maximize Z=  V j X j
j 1

Subject to
n

W
j 1
j X j K

1, if object j is selected

X j 
 0, otherwise
Example – Investment Planning
 The board of directors of General Wheels Co. is
considering seven large capital investments. Each
investment can be made only once.
 These investments differ in the estimated long-run
profit (NPV) that they will generate as well as in the
amount of capital required, as shown in the next
slide (in units of millions of dollars)
Investment Planning example…

Investment Opportunity

1 2 3 4 5 6 7

Estimate 17 10 15 19 7 13 9
d Profit
Capital 43 28 34 48 17 32 23
required
Investment Planning example…
 The total amount of capital available for these investments is
$100 million.
 Investment opportunities 1 and 2 are mutually exclusive, and
so are 3 and 4.
 Furthermore, neither 3 nor 4 can be undertaken unless one of
the first two opportunities is undertaken.
 There are no restrictions on investment opportunities 5,6 and
7.
 The objective is to select the combination of capital
investments that will maximize the total estimated long-run
profit.
 Set Covering Problem
 A telecommunication company is considering expanding
its mobile phone service operations into a new area.
 The area is divided into 10 neighborhoods.
 The company is considering 5 locations (to set up mobile
phone towers) to reach all 10 neighborhoods.
 It costs a significant amount of money to set-up the facility
at a node and make it operational
 The company would like to minimize total costs of setting-
up the towers but at the same time reach all of the
neighborhoods.
Set Covering
Problem Node1 : 1, 2, 7, 8, 9
Node3 : 1, 2, 3, 4, 5, 10
2 3 Node5 : 3, 4, 5, 6, 9, 10
Node7 : 1, 5, 6, 7, 8, 9, 10
1 10 Node9 : 3, 6, 7, 8, 9,10
9
4 Cost of opening
5 nodes:
8 5
6 Node 1: 125
Node 3: 85
7 Node 5: 70
Node 7: 100
Node 9: 110

Determine which nodes should be opened to

provide coverage to all neighborhoods at a
minimum cost?
 Service Nodes
1 3 5 7 9
Cost 125 85 70 100 110
1 1 1 0 1 0
2 1 1 0 0 0
3 0 1 1 0 1
Localitie
4 0 1 1 0 0
s
5 0 1 1 1 0
6 0 0 1 1 1
7 1 0 0 1 1
8 1 0 0 1 1
9 1 0 1 1 1
10 0 1 1 1 1
Mathematical Model
1, if node j provides service (j = 1,3,5,7,9)
Xj 
 0, otherwise
Minimize Z=125X 1  85 X 3  70 X 5  100 X 7  110 X 9
Subject to
X 1  X 3  X 7 1
X 1  X 3 1
X 3  X 5  X 9 1
X 3  X 5 1
X 3  X 5  X 7 1
X 5  X 7  X 9 1
X 1  X 7  X 9 1
X 1  X 5  X 7  X 9 1
X 3  X 5  X 7  X 9 1
General form
n
Minimize Z=  C j X j Applications:
j 1
1. Vehicle routing problems
Subject to
n 2. Facility location
a
j 1
ij X j 1  i 1.....m
3. Airline crew scheduling

X j   0,1 4. Circuit design

5. Resource allocation
6. Capital investment
Fixed Charge Transportation
Problem
SUPPL DEMAND
Y
S1 1 C11
C12
C14 C13 1 D
1
S2 C21
2 C22
C23
C24 2 D
2
S3 3

3 D
3
S4 4
C51 C52
4 D
C53 4
S5 5 C54
Fixed Charge Transportation
Problem
Variable cost of transportation
1 2 3 Supply
1 6 4 9 10
2 8 5 4 20
3 10 7 3 30
Demand 20 20 20

Fixed cost of transportation

1 2 3
1 500 1000 700
2 600 700 800
3 200 400 900
Fixed Charge Transportation
Problem
Notations:
m- Number of sources
n- Number of destinations
Si - Supply at source i, i = 1, 2, 3, 4,5

Dj - Demand at destination j, j = 1, 2, 3, 4

FCij - Fixed cost of transportation between

source i and destination j
VCij - Variable cost of transportation per unit from
source i to destination j
Fixed Charge Transportation
Problem m n
Minimize Z    FCijYij  VCij xij 
i 1 j 1

Subject to
n

x
j 1
ij  Si , i 1, 2,...m

x
i 1
ij Dj , j 1, 2,...n

xij MYij ,  i 1, 2,...m and j 1, 2,...n

xij 0  i 1, 2,...m and j 1, 2,...n
1, if shipment takes place between source i and destination j
Yij 
 0,otherwise
Facility Location Problem
Decision 1:
Hyderaba Where to locate the
d 1 D1
facilities?
Mumbai Kochi
Decision 2:
Chennai Surat From each chosen facility, 2 D2
how much should be
Delhi Lucknow
shipped to the destinations
Bangalore 3 D3
Pune for satisfying the demand?
Kolkata Vizag
Costs incurred:
1. Fixed cost of locating a
Ahmedaba 4 D4
d facility (Fi)
2. Transportation costs (Cij)
Set of Potential
locations for Destinations
facilities (m) (n)
 Facility Location
Problem…
Facilit Fixed Variable cost of transportation
y cost of
settin
1 2 3 4 Supply
g up
facility
1 9 6 2 5 7500
1 100000
2 4 9 8 2 5000
2 50000
3 6 8 5 9 10000
3 75000
4 7 3 4 8 6000
4 48000
5 2 4 5 7 3000
5 62000
6 5 7 9 3 4500
6 84000
7 3 8 2 4 2500
7 70000
Deman
6000 8000 5000 4500
d
Facility Location Problem
m m n
Minimize Z=  Fi X i    Cij yij
i 1 i 1 j 1

Subject to
n

y
j 1
ij Si X i  i 1.......m

y
i 1
ij  D j  j 1.......n

1, if location i is selected

X i 
0, otherwise
yij  Quantity shipped from facility i to destination j
Si  Supply at facility i
D j  Demand at destination j
Good Products Company
 The Research and Development division of the GOOD
PRODUCTS COMPANY has developed three possible new
products. However, to avoid undue diversification of the
company’s product line, management has imposed the
following restrictions.

 From the three possible new products, at most two should

be chosen to be produced

 Just one of the two plants should be chosen to be the sole

producer of all the new products.
Good Products Company…

Production time used for

each unit produced Production time
available per
Product Produc Product week
1 t2 3

Plant 1 3 hours 4 hours 2 hours 30 hours

Plant 2 4 hours 6 hours 3 hours 40 hours

(thousands of
Unit Profit 8 7 6
dollars)
Sales
5 7 9 (Units per week)
Traveling Salesman Problem
(TSP)
 Problem Description:

Given a list of cities and their pair-wise distances, the

problem is to find the shortest possible tour for a

salesman who starts from a city, visits each and every

city exactly once and comes back to the starting city

 This problem is also known as Hamiltonian Circuit in

Graph Theory.
Interesting facts about TSP
 Number of feasible solutions for a ‘n’ city problem?

 For a 10 city problem:36,28,800 solutions

 For a 15 city problem:1,307,674,368,000!!!!!

 Optimal solution for 15,112 towns in Germany (2001)

- Network of 110 processors at Rice and Princeton

- Computational time equivalent: 22.6 years

 Optimal solution for 33,810 points on a PCB (2005)

 85,900 points – Concorde TSP solver - current record – April

2006 (136 CPU years)
Applications of TSP
 Logistics: Vehicle routing

 PCB manufacturing – Soldering using Robots

 Gene mapping – DNA sequencing

 Protein function prediction

Mathematical model of TSP
n n
Minimize Z   Cij X ij
i 1 j 1

Subject to
n

X
j 1
ij 1, i 1, 2,...n

X
i 1
ij 1, j 1, 2,...n

1, if travelled from city i to city j

X ij 
0, otherwise
Cost of traveling from city i to city j (or )

Cij  Distance from city i to city j (or)
 Travel time between city i to city j

Sub-tour Elimination???

OR
• Miller-Tucker-Zemlin (MTZ) Formulation (1960)

ui  u j  1 ( N  1)(1  X ij ) i 1, j 1
u1 1, 2 u j  N ,  j 2,..., N
The Seven Bridges of
Königsberg, Germany
 In the city of Konigsberg, there was a park through
which a river ran. In this park seven bridges crossed
the river. The problem at the time was to plan a walk
that crossed each of the bridges exactly once, starting
and ending at the same point.
1

2 3

4
Eulerian Circuit or Eulerian
Tour

An Eulerian circuit, in a graph is a cycle that

uses each edge precisely once. If such a cycle

exists, the graph is called Eulerian

1 Redrawn 2
Graph:

2 3 4 1
3
4
 Applications – Eulerian
circuit
Applications:
1. Trash Collection
2. Transmission line inspections
3. Road-sweeping
4. Snow-plows
5. Highway lawn movers
6. Police patrolling
Chinese Postman Problem
 The problem of determining the minimum
additional weights(edges) to be added to a
non-Eulerian graph to make it Eulerian
Solution methodologies for IP

 P, NP, NP-Hard, NP-Complete problems

 Complete enumeration
 Branch and Bound
 Branch and Cut
 Branch and Price
 Heuristics
 Meta-Heuristics: GA, SA, ACO, TS
SUDOKU

Rules:
1. Use numbers 1,2…..9
2. There should not be
repetition in any row or
column
3. Each box should have
all the numbers 1,2…9
Some more interesting facts about
SUDOKU
 Sudoku – Hward Garnes - 1979 - U.S
 Creation of SUDOKU – brute force
 The number of SUDOKU matrices for the standard 9x9 game
is:
6,670,903,752,021,072,936,960
= 6.67 x 10 21

 Properly posed SUDOKU – Unique Solution

 Number of givens does not matter
 Position of the givens matter
 Number of givens to generate an unique solution
 NP-Complete
Graph Coloring
Problem
Your mission is to color
the entire map of South
America
1. No country may touch
another country of the
same color
2. You will be charged
each time you use a
new color
3. You must color the
map as cheaply as
possible
Graph Coloring
Problem
Your mission is to color
the entire map of South
America
1. No country may touch
another country of the
same color
2. You will be charged
each time you use a
new color
3. You must color the
map as cheaply as
possible
Applications

 Scheduling: Assigning jobs to machines to minimize

the make-span
 Assigning aircraft to flights
 Bandwidth allocation to radio stations
 Pattern matching
SUDOKU: 81 node graph coloring
problem
 Mathematical model
1 if 'k' occupies the cell (i x j) in a 'n x n' matrix
X ijk  
0 otherwise

X
i 1
ijk 1, j 1....n, k 1.....n

X
j 1
ijk 1, i 1....n, k 1.....n
 Sub-matrix constraint

mq mp

 
j mq  m 1 i mp  m 1
X ijk 1, k 1....n, p 1.....m, q 1....m

Reference: Andrew C. Bartlett and Amy N. Langville, 2006

Military Logistics
(1/3)
 Moving men and materials in large numbers and quantities
 Land forces deployed as convoys
 Vehicles must travel nose to tail in a pre-specified order
 Peace time: Convoys allowed to halt at pre-decided
locations; coordinate with civilian transport agencies
 War time: Highest priority over other users of road / rail;
 Non-stop travel until destination
Military Logistics (2/3)

 Minimum headway
 No-Crossing: Applicable for both peace time and
wartime
Reasons:
 Roads not having load bearing ability and width
to accommodate two convoys at a time
 To avert accidents
 To avoid mix-up/confusion
Military Logistics (3/3)

Decision Peace Time War Time

Stoppage Can halt at No stoppage

en-route nodes/cities en-route

Priority All convoys Priorities varies

have equal over time
priority
Crossing/ Not permitted Not permitted
overtaking of
convoys
Artillery Tractors

Armored Combat Tanks

 Rocket Artillery
Tanks
 Infantry Tanks

Armored Personnel carriers

Convoy Movement Problem
The Convoy Movement Problem (CMP) can be defined
as
 The problem of minimizing the travel time/travel
span of military convoys from their home bases to
their designated destinations
 Operating on a multimodal transportation network
 While obeying constraints on scheduling and routing
issues as well as
 The availability and use of various types of
transportation assets (TAs)
Instances of CMP
“Theory without application is useless and application
without theory is dangerous”
- Anonymous
 Scheduling trains on single line tracks
 Routing of Automated Guided Vehicles (AGV) in Flexible
Manufacturing Systems (FMS) environment
 Routing of baggage from different flights along a
common automated transportation system in airports
 Hazardous material transportation
 Emergency response management
Integer Programming (IP)
Model for Convoy
Movement Problem
 IP model
(1/7)
The mathematical model developed for CMP is valid under the
following assumptions
 Convoys operate on single mode of transportation
 Convoys move at same speed across the whole network
 All convoys have equal priority
 All the nodes have sufficient infrastructure for halting
more than one convoy
 Infinite and continuous availability of transportation
assets
 All convoys are ready at their respective home bases at
time 0
 IP model (2/7)

 Sets and Indices

Q Set of all the nodes

A Set of arcs
S Set of source nodes
D Set of destination nodes
C Set of military convoys need to be
moved
M very large integer
 IP model (3/7)
ATqc  Arrival time of convoy ' c ' at node ' q ', q  S
DTqc Departure time of convoy ' c ' at node ' q ', q  D

1 if convoy c traverses between nodes q and q'

A 
c
qq '
0 otherwise

1 if convoy c traverses between nodes q and q' before convoy c'


X qqcc ''  if convoy c' traverses between nodes q and q' before convoy c
0  if arc qq' is not on the path of either convoy c or c' or both
 

1 if convoy c traverses between nodes q and q' before convoy c' traverses between nodes q' and q

Yqqcc' '   if convoy c' traverses between nodes q' and q before convoy c traverses between q and q'
0  if arc qq' is not on the path of convoy c or c' or both
 
 IP model (4/7)

Objective Function

Minimize  AT c
c
qd
c
 IP model
1. Network Flow Conservation (5/7)
Constraints
1 if q qsc 
 
 Aqq '   Aq '' q 0 if q qs or qd   c  C , q q ', q q '' and (q, q ', q '')  Q
c c c c

q' q ''  c 
  1 if q  q d 
2. Arc Travel time Constraints
ATqc'  M (1  Aqq
c
' ) T A c
qq ' qq '  DT c
q 

c c c c 
 c  C , (q, q ')  Q and q q '
ATq '  M (1  Aqq ' ) Tqq ' Aqq '  DTq 

3. Halt enabling constraint

DTqc  ATqc 0, c  C , q  S , and q  D
 IP model …(6/7)

4. Minimum Headway constraint for convoys

traveling along the same arc in the same direction

MX qqcc ''  M (1  Aqqc ' )  M (1  Aqqc ' ' )  ATqc'  ATqc' '  h(1  X qqcc '' )  (c, c ')  C , (q, q ')  Q,
cc ' 

M (1  X qq ' )  M (1  Aqq ' )  M (1  Aqq ' )  ATq '  ATq '  hX qq '  c c ', q q ' and c '  c
cc ' c c' c' c

5. Binding Constraints

X qqcc ''  Aqqc ' (c, c ')  C ,(q, q ')  Q, c c ', q q ' and c '  c
X qqcc ''  Aqqc ' ' (c, c ')  C ,(q, q ')  Q, c c ', q q ' and c '  c
Yqqcc''  Aqqc ' (c, c ')  C , (q, q ')  Q, c c ', q q ' and c '  c
Yqqcc''  Aqc''q (c, c ')  C , (q, q ')  Q, c c ', q q ' and c '  c
 IP model …(7/7)
6. No-crossing
constraints
MYqqcc''  M (1  Aqqc ' )  M (1  Aqc''q )  DTqc  ATqc '  (c, c ')  C (q, q ')  Q
c

M (1  Yqq ' )  M (1  Aqq ' )  M (1  Aq ' q )  DTq '  ATq '  c c ', q q ', and c '  c
cc ' c c' c'

7. Boundary conditions

ATqcc 0, c  C and qsc  S

s

DTqcc M , c  C and qdc  D

d
Mathematical model for CMP
 Proposed model can solve problem instances up to
50 cities and 10 convoys within 8 hours of
computational time
 For a 50 node, 10 convoy problem, with an arc
density of 0.25:
Approximate No. of variables : 32,000
Approximate No. of constraints : 1,50,000
 Beyond 50 cities, lower bounds can be developed
and used for generating solutions
Lagrangean Relaxation (LR) based
lower bound
 In Lagrangean Relaxation, constraint sets (4) and (6)
are relaxed and included in the objective function with
Lagrangean multipliers
 The initial values for Lagrangean multipliers were
generated randomly in the range of 1 – 2
 Subgradient optimization procedure is used to tighten
the lower bound
 LR (PT_CMP) Problem
 qq ' qq '  q q ) 
q
T
q'
Ac
 ( DT c
 AT c

c q

 qqcc ''      ATqc' '  ATqc'  h(1  X qqcc '' )  MX qqcc ''  M (1  Aqqc ' )  M (1  Aqqc ' ' )  
cC c 'C qQ q 'Q

 qqcc''      ATqc'  ATqc' '  hX qqcc ''  M (1  X qqcc '' )  M (1  Aqqc ' )  M (1  Aqqc ' ' )  
cC c 'C qQ q 'Q

qqcc ''      ATqc '  DTqc  MYqqcc''  M (1  Aqqc ' )  M (1  Aqc''q )  

cC c 'C qQ q 'Q

qqcc ''      ATqc'  DTqc' '  M (1  Yqqcc'' )  M (1  Aqqc ' )  M (1  Aqc''q ) 

cC c 'C qQ q 'Q

Subject to constraints (1) – (3),(5), and (7)