Integer Programming

Decision Variables take Integer Values

X2

Area of Feasibility 5 4 3 2 1

Point at Intersection are feasible points as they are integers Acceptable values For X1, X2 are (1,1), (2,1), (1,2), (2,2) ..

X1

A cost minimization example

Minimize Cost = 0.01X1 + 0.07X2 Subject to : 6X1 + 2X2 >= 18 (Constraint 1) 8X1 + 10X2 >= 40 (Constraint 2) X2 >= 1 (Constraint 3) X1, X2 >= 0 (Non Negativity)

X2
Minimization Problem 10 A 8

Z = 0.10x2.27 + 0.07x2.19 = 0.38 Z (Integer) = 0.10*2 + 0.07*3 = 0.41

Objective Function 4 3 2

B
C 2 4 6 8 X2 >= 1 10 X1

Optimum X1 = 2.27, X2 = 2.19 Optimum (Integer) X1 = 2, X2 = 3

Another Example
Maximize Z = 40000X1 + 30000X2 Subject to: 10X1 + 3X2 <= 30 600X1 + 1000X2 <= 3600 500X1 + 200X2 <= 1800 X1,X2 >= 0 and integers Integer : X1 = 2, X2 = 2, Z = 140000 LP : X1 = 2.561, X2 = 1.4634, Z = 146341.46 Simply cannot round LP solutions to get Integers

Types of Integer Programming Problems

Pure Integer Problems (Number of each type of plane to produce, number of each type of house to construct, and so on..) Mixed Integer Problems (have some variables requiring integer values, and some may have continous (decimal) values, like a landscape design which requires the percentage area to plant grass (continous variable) and number of trees to plant (integer variable) 0-1 Integer Problems where variables require values as 0 or 1 (binary), or Yes/No type. Examples are a bank considering possible locations for a branch, scheduling of jobs to machines or workers to jobs.

Formulating Integer programming Problems with 0-1 constraints

Either - or Alternatives k out of n alternatives If Then Alternatives Either Or Constraints Variables that have minimum level requirements

Either Or Alternatives
A manufacturer may need a machine to replace one that recently has failed. Two alternatives X1 and X2 are being considered, but only one will be needed. Constraint : X1 + X2 = 1 If neither machine will be acquired then constraint is X1 + X2 <= 0 X1 and X2 are 0 or 1 variables

K out of N Alternatives
A decision maker (DM) must choose a specified number of alternatives. Say choose 2 machines from a list of 5 alternatives. X1 + X2 + X3 + X4 + X5 = 2 (exactly 2) X1 + X2 + X3 + X4 + X5 >= 2 (atleast 2) X1 + X2 + X3 + X4 + X5 <= 2 (atmost 2) Choosing anywhere 2 to 4 X1 + X2 + X3 + X4 + X5 >= 2 (lower bound) X1 + X2 + X3 + X4 + X5 <= 4 (upper bound)

If Then Alternatives
DM has to take some action which necessitates another action that supports the initial decision Purchase of X2 machine may necessitate purchase of another X1 machine Purchase X2 (X2=1) will lead to purchase of X1 (X1=1). If X2 is not purchased, (X2=0) will lead to X1 not being purchased (X1=0) X1 >= X2 or X1 X2 >= 0 (Reverse is not true Purchase of X1 does not lead to purchase of X2) If purchase of either machine requires the purchase of the other then X1 X2 = 0

Either Or Constraints
Situations may arise in which a constraint will apply only if a particular alternative is chosen A certain machine may necessitate special power requirements. Hence it can be able to turn-on or turn-off a constraint. A machine X3 requires the constraint 5X1 + 3X2 >= 100. Formulate as 5X1 + 3X2 >= 100X3, X3 is 0-1 variable 5X1 + 3X2 100X3 >= 0 If X3 is not chosen, constraint 5X1 + 3X2 >= 50 is required. Formulate as : 5X1 + 3X2 >= 100X3 (if X3 is chosen) 5X1 + 3X2 >= 50(1-X3) (if X3 is not chosen)

Either Or Constraints Choice of X3

If X3 is ON (X3 = 1) 5X1 + 3X2 >= 100 Formulate as 5X1 + 3X2 >= 100X3, X3 is 0-1 variable 5X1 + 3X2 100X3 >= 0 X3 is 0-1 variable If X3 is OFF (X3 = 0) 5X1 + 3X2 >= 50 Formulate 5X1 + 3X2 >= 50(1-X3) (if X3 is not chosen) X3 is 0-1 variable

Variables that have Minimum Level Requirements

At times, a variable either will have a zero value or an amounts that exceeds a specified value. Example is the minimum order size for a purchased part might be required by a vendor. Say min qty for X1 is 200 units. X1 >= 200Y1 (Y1 is 0-1 variable)
X1 200Y1 >= 0, X1 = integer, Y1 = 0 or 1 If X1 = 0, the constraint would force Y1 to be zero

Specialized Integer Programming Problems

Fixed Charge problem Set Covering problem Knapsack problem Facility Location problem Traveling Salesperson problem

Fixed Charge Problem

A company makes 3 products A,B,C. Unit profit for A,B,C are $6,$10,$5 respectively. The products can be manufactured using one of the two processes. Demand for A is predicted to be between 50 and 100 units per week. For product B, demand is between 150 to 200 units per week. Similarly for product C, it is predicted to be between 100 to 150 units per week. Process 1 has a capacity of 2000 hours per week, with product A, B, C taking 4 hours, 6 hours and 3 hours per unit. The setup cost for process 1 is $100 and setup time is 24 hours. Process 2 has a capacity of 2400 hours per week, with product A, B, C taking 5 hours, 7 hours and 4 hours per unit. The setup cost for process 2 is $80 and setup time is 18 hours. Determine the production schedule that will maximize the profit and also determine which of the two processes will be utilized.

Fixed Charge Problem

Let X1, X2, X3 be units of Product A,B, C Let Y1, Y2 be the 0-1 variables for Process 1 and 2 Maximize Z = 6X1 + 10X2 + 5X3 100Y1 80Y2 Subject to 4X1 + 6X2 + 3X3 + 24Y1 <= 2000 hours (Process 1) 5X1 + 7X2 + 4X3 + 18Y2 <= 2400 hours (Process 2) X1 >= 50 X1 <= 100 X2 >= 150 X2 <= 200 X3 >= 100 X3 <= 150 Y1 + Y2 = 1 (Any one Process is to be selected) X1,X2,X3 >= 0 and integer Y1,Y2 = 0 or 1 Optimal Solution X1 = 80, X2 = 200, X3 = 150, Y1 = 1, Y2 = 0, Max Profit = $3330

Process 1

Any one Process to be selected

Process 2

Incur Set up Cost And Set up Time

Set Covering Problem

A telecommunication company is considering expanding its cable and internet services operations into a new area. The area is divided into 10 neighborhoods. The company is considering 7 location nodes to reach all 10 neighborhoods The cost for opening the seven nodes are $125, $85, $70, $60, $90, $100 and $110. Seven nodes can reach the following neighborhoods : Node 1 : Neighborhoods 1,3,4,6,9, 10 Node 2 : Neighborhoods 2,4,6,8 Node 3 : Neighborhoods 1,2,5 Node 4 : Neighborhoods 3,6,7,10 Node 5 : Neighborhoods 2,3,7,9 Node 6 : Neighborhoods 4,5,8,10 Node 7 : Neighborhoods 1,5,7,8,9 Determine which nodes should be opened to provide coverage to all neighborhoods at a minimum cost

Neighbourhood

1
1 2 3 4 5 6 7 Y

Node

Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y

1 0 Y Y

Set Coverage Problem

Let Cj be cost of operating or setting up node j (j = 1,2,3,7) Let Xj = 1 if node j provides service (j = 1,2,3,7) Let Xj = 0 if node j does not provide service (j = 1,2,3,7) Minimize Z = 125X1 + 85X2 + 70X3 + 60X4 + 90X5 + 100X6 + 110X7 Subject to : X1 + X3 + X7 >= 1 (coverage of neighborhood 1) X2 + X3 + X5 >= 1 (coverage of neighborhood 2) X1 + X4 + X5 >= 1 (coverage of neighborhood 3) X1 + X2 + X6 >= 1 (coverage of neighborhood 4) X3 + X6 + X7 >= 1 (coverage of neighborhood 5) X1 + X2 + X6 >= 1 (coverage of neighborhood 6) X4 + X5 + X7 >= 1 (coverage of neighborhood 7) X2 + X6 + X7 >= 1 (coverage of neighborhood 8) X1 + X5 + X7 >= 1 (coverage of neighborhood 9) X1 + X4 + X6 >= 1 (coverage of neighborhood 10) X1, X2, X3, X4, X5, X6, X7 = 0 or 1 Optimal Answer X2 = 1, X4 = 1, X7 = 1 Cost = $255

KnapSack Problem
The Knapsack problem is defined as how many units of different kinds of items or products to put in a knapsack with a given capacity in order to maximize profit. Maximize Z = C1X1 + C2X2 + . + CnXn Subject to : A1X1 + A2X2 + .. + AnXn <= B X1, X2,.. Xn >= 0 and integer

Knapsack Example
TriState must purchase three possible major food items : poultry, ice-cream, yoghurt in batches to realize quantity discounts for filling up vacant 5 tons capacity in its super freezer. One batch of ice-cream weighs 2 tons and profit per batch is $90. One batch of poultry weights 3 tons and profit per batch is $150, while one batch of yoghurt weighs 1 ton and profit per batch is $30. Let X1,X2,X3 = batch of ice-cream, poultry, yoghurt Maximize Z = 90X1 + 150X2 + 30X3 Subject to : 2X1 + 3X2 + X3 <= 5 tons X1, X2, X3 >= 0 and integer Optimal Solution : X1 = 1, X2 = 1, Max Profit = $240

Facility Location Problem

Here we consider many new locations and also take capacity considerations 0-1 mixed integer programming is utilized. Used for location of plants, hospitals, healthcare facilities, fast food restaurants, schools, police and fire stations.

Facility Location Problem

M&R Manufacturing Company is considering a major expansion. It has 2 plants A & B with capacity of 26000 units and 30000 units per annum respectively. The company sells through its retail outlets located at P, Q, R and S with annual demands of 27000, 32000, 23000 and 30000 units per annum. The company is in the process of considering four new manufacturing plant locations at I, J, K and L. The capacities and cost of purchasing or building the facility is as : I 30000 units / $220000, J 33000 units / $260000, K 26000 units / $ 200000, L 37000 units / $ 280000. The shipping costs from the existing plants and the 4 new plant locations to the existing retail outlets is given. Determine the optimal shipping schedule that minimizes the total cost which includes the shipping cost and cost of building/purchasing, such that the total demand at all retail outlets is satisfied.

Shipping Cost per Unit Plant A RO P 7 RO Q 5 RO R 4.5 RO S 5.5 Supply 26000

B I (New) J (New)

5 9 6

7 6.5 3.5 6 4.5 32000

12 2 5 2.5 5 23000

11 3.5 3 4 3 30000

30000 30000 33000 26000 37000

K (New) 8 L (New)
Demand

6.5 27000

Total Demand = 112000, while Existing Capacity = 56000. New Capacity to be added should be 56000 or more

Facility Location Problem

m = number of sources of supply (m = 1,2) k = number of sources of new supply (k = 1,2,3,4) j = number of destinations (j = 1,2,3,4) Cij = Shipping cost of one unit from source i to destination j Xij = Units shipped from source I to destination j Pk = Purchase cost of plant k Min Z = CijXij + PkYk for all i, j and k Subject to : Xij <= Si for all i (Supply at existing plant, I = 1,2) Xij <= SiYifor all i (Supply at new proposed plant, i = 3,4,5,6) Xij <= Dj for all j (Demand at destination, j = 1,2,3,4) Xij >= 0, Yk = 0 or 1

Shipping Schedule Plant A RO P RO Q 26000 RO R RO S Supply 26000

B I (New) J (New) K (New) L (New)

Demand

27000

3000 30000

30000 30000 33000

3000

23000

26000 37000

27000

32000

23000

30000

112000

Total Demand = 112000, while Existing Capacity = 56000. New Capacity to be added should be 56000 and added with plant I and K The cost of building or Purchasing plant I and K is 220000 + 200000 = $420000 The total Minimum Cost is $ 886500. Shipping Cost = $466500

Traveling Salesperson Problem

It attempts to minimize total cost, distance or time of departing location I and returning to same location I in a tour, visiting all locations once. It is like the transportation problem, but relatively complicated.

Traveling Salesperson Example

The City Garbage collection team wants to determine the best route for their garbage trucks. There are 4 sections in the city under consideration. The time of travel (in minutes) between the various sections of the city is given. If location 1 is the garage from which garbage trucks leave and to which they return, formulate this problem as a traveling salesperson problem

Travel Time (in minutes) between locations

From/ To 1
2 3 4

1
0 8 6 5

2
8 0 3 7

3
6 3 0 9

4
5 7 9 0

Traveling Salesperson Example

Let Xijk = 0 or 1 for travel from start location i to end location j in journey leg k. There are 4 locations : 1, 2, 3, 4 In leg 1, start from node 1 and go to node 2, node 3, node 4 In leg 4, come to node 1 from node 2, node 3, node 4. In leg 2, go from node 2 to node 3, node 4 or go from node 3 to node 4 or node 2 In leg 3, go from node 2 to node 3, node 4 or go from node 3 to node 4 or node 2

Various Routes when started from Node 1

Start

Leg 1

Leg 2

Leg 3

Leg 4

1
1 1 1 1 1

2
2 3 3 4 4

3
4 2 4 3 2

4
3 4 2 2 3

1
1 1 1 1 1

Decision Variables (DV) for each leg in a 4 node problem

Leg DV DV DV DV 1 2 3 4 X121 X131 X141 1

2 3 4
X232 X233 X214 X242 X243 X314 X342 X343 X414 X422 X423

DV 5

DV 6

X432 X433

X322 X323

Xijk = travel from start location i to end location j in journey leg k.

Constraints
For every leg of journey (1, 2, 3,and 4), no of journeys = 1 No of Visits from Node 2, 3 and 4 = 1 No of Visits to Node 2, 3 and 4 = 1 Leg 2 starts where Leg 1 ends for Node 2, 3 and 4 Leg 3 starts where leg 2 ends for Nodes 2, 3 and 4 Leg 4 starts where leg 3 ends for Nodes 2, 3 and 4

Problem Formulation
Minimize Z1 = 8X121 + 6X131 + 5X141 Minimize Z2 = 3X232 + 7X242 + 9X342 + 7X422 + 9X432 + 3X322 Minimize Z3 = 3X233 + 7X243 + 9X343 + 7X423 + 9X433 + 3X323 Minimize Z4 = 8X214 + 6X314 + 5X414 Minimize Z = Z1 + Z2 + Z3 + Z4 Subject to : X121 + X131 + X141 = 1 (Starting node is 1 or Leg 1 visits limited to 1) X214 + X314 + X414 = 1 (Ending Node is 1 or Leg 4 visits limited to 1) X232 + X242 + X342 + X422 + X322 + X432 = 1 (leg 2 visits limited to 1) X233 + X243 + X343 + X423 + X323 + X433 = 1 (leg 3 visits limited to 1) X214 + X232 + X242 + X233 + X243 = 1 (visits from node 2 are limited to 1) X314 + X322 + X323 + X342 + X343 = 1 (visits from node 3 are limited to 1) X414 + X422 + X432 + X423 + X433 = 1 (visits from node 4 are limited to 1) X121 + X322 + X422 + X323 + X423 = 1 (visits to node 2 are limited to 1) X131 + X232 + X233 + X432 + X433 = 1 (visits to node 3 are limited to 1) X141 + X242 + X243 + X342 + X343 = 1 (visits to node 4 are limited to 1) X121 = X232 + X242 (leg 1 ends where leg 2 starts) X131 = X322 + X342 (leg 1 ends where leg 2 starts) X141 = X422 + X432 (leg 1 ends where leg 2 starts) If leg 2 of trip ends at location 2,3 or 4, leg 3 of trip must start at node 2, 3 or 4 X322 + X422 = X233 + X243 (Node 2) X232 + X432 = X323 + X343 (Node 3) X242 + X342 = X423 + X433 (Node 4) X323 + X423 = X214 (last leg of trip must end at location 1) X233 + X433 = X314 ( --- do ---) X243 + X343 = X414 (---- do ---) Xijk = 0 or 1 for all I,j,k

Optimal Solution
Travel Time (in minutes) between locations

From/ To 1
2 3 4

1
0 8 6 5

2
8 0 3 7

3
6 3 0 9

4
5 7 9 0

Total Time = 6 + 3 + 7 + 5 = 21 Route is location 1 to 3 to 2 to 4 to 1

Difficulties in Solving Integer Programming Problems

In Integer Programming, a finite number of integer points are considered which are more than the feasible points in linear programming. As the number of variables (n) increase, the no of solutions are 2n in 0-1 programming. In pure integer programming, the number of solutions taking integer values are still more as compared to 0-1 variables in 0-1 programming. The computational procedures for integer programming are not as simple as that in linear programming.