Sunteți pe pagina 1din 5

CBA 2019: Assignment II

Due: April 21st, 2019

Instructions: This is a group assignment, and only one final write-up and GAMS models must be
submitted per group. Submit your assignment (including all GAMS model files) in a “.zip folder”,
with the name of the folder “yourlastnames_hw2”, on the course website.

1. Consider the following Fixed Charge Location Problem (FCLP) discussed in class. Stripe
Shoes, Inc. must build a set of production facilities to meet demand. There is a set 𝑖 ∈ {1, … , 𝐼}
of potential facility locations. Opening location 𝑖 costs 𝑓𝑖 dollars and, if opened, can produce a
maximum of 𝑢𝑖 units of shoes. The demand can only be met by shipping shoes from an open
facility to regional warehouses used for distribution. The regional warehouses are indexed by
the set 𝑗 ∈ {1, … , 𝐽}. Each regional warehouse 𝑗 is required to receive 𝑏𝑗 units of shoes. The
cost of shipping one unit of shoes from facility 𝑖 to warehouse 𝑗 is 𝑐𝑖𝑗 .

Table 1. Data for the FCLP

Warehouses Building Upper


1 2 3 4 5 6 7 8 9 10 11 12 13 14
Facilities Costs Bounds
1 1 4 10 9 8 1 10 7 2 3 10 10 3 3 85 188
2 9 7 3 2 10 6 7 1 9 3 1 8 8 7 65 182
3 9 1 9 6 6 1 1 9 8 8 5 1 1 9 94 141
4 4 5 8 7 3 5 8 10 2 10 1 4 5 9 77 170
5 8 1 3 3 10 9 10 2 2 6 1 3 4 1 67 137
6 2 3 5 9 4 2 6 2 7 5 1 4 3 5 93 181
7 9 4 3 6 4 4 4 4 6 6 7 8 3 9 75 140
8 10 5 1 9 1 3 5 10 10 3 10 8 9 3 75 192
9 2 9 9 10 3 8 2 1 6 4 9 10 8 5 68 182

Demands 9 1 7 8 2 8 1 1 1 4 10 2 1 1

(a) Test the strength of Formulations 1 and 3 discussed in class by solving this problem using
GAMS with the data given in Table 1. Report the LP-IP gap for the various formulations
tested, and state which is a better model and why.
(b) Suppose that if facility 1 is built, then facility 2 must also be built. What constraint will
you need to add to the model? Resolve the original problem with this constraint added.
(c) Suppose that at least 3 facilities have to be built. What constraint will you add to the model
to enforce this condition? Resolve the original problem with this condition added.
(d) Suppose that all the demand at a warehouse must be satisfied from a single production
facility. (This is referred to as single sourcing.) Formulate this problem as an integer
program, clearly defining all the decision variables and constraints, and resolve the
problem.
2. There are seven important shopping regions in the Hyderabad area: Ameerpet, Begumpet,
Charminar, Dilsukhnagar, Erragadda, Falaknuma, and Gachibowli (A, B, C, D, E, F, and G for
short). The Greater Hyderabad Municipal Corporation (GHMC) wants to achieve the goal that
every person living in any one of these regions has to be within 10km of a “lifestyle mall”.
You have been brought in as an analytics consultant to decide where to build these malls. There
are proposals to build a mall in each of the seven regions. The distances between the various
regions and costs of building a mall in each region are given in Table 2 below.

(a) Formulate an integer program that will determine the locations to build lifestyle malls at a
minimum cost subject to the constraint that each district is within 10km of a mall.
(b) Solve the problem using GAMS with the data given in Table 2, and report which areas will
get selected in the optimal solution.
(c) The MLAs of Begumpet and Gachibowli are at loggerheads, and so, the MLA of
Gachibowli has said that if a mall is located in Begumpet then a mall will not be located in
Gachibowli. Modify your model to accommodate this constraint and resolve the problem.
Report which areas will have malls built in the optimal solution.
(d) The MLAs of Begumpet and Gachibowli have made up with each other; so, the constraint
of part (c) does not hold anymore. Now suppose that if a mall is located in Begumpet or
Gachibowli (or both), then a mall has to be located in one of the other regions. What is the
new set of malls that will be built?
(e) It is now learnt that the MLAs of Begumpet and Gachibowli reconciled via negotiations
with the MLA of Erragadda. So, now if a mall is located in Begumpet or Gachibowli (or
both), then a mall has to be located in Erragadda. Accommodate this constraint and resolve
the problem. Did reconciliation help either of these MLAs in this case?

Table 2. Distances between different regions and cost of building malls in each region

Regions A B C D E F G Costs
A 0.0 6.1 15.2 17.0 16.8 8.4 16.6 10
B 6.1 0.0 7.6 16.0 11.3 2.2 11.9 6
C 15.2 7.6 0.0 22.0 17.3 10.2 16.7 7
D 17.0 16.0 22.0 0.0 12.1 14.8 7.2 8
E 16.8 11.3 17.3 12.1 0.0 9.4 2.9 13
F 8.4 2.2 10.2 14.8 9.4 0.0 9.2 9
G 16.6 11.9 16.7 7.2 2.9 9.2 0.0 8
3. Solve the following problem using the branch-and-bound method. Clearly show the entire
tree, including values of decision variables and objective function value at each node. (You
can use GAMS to solve the LP relaxations at each node.)

Maximize 10𝑥1 +  2𝑥2 +  7𝑥3   +  5𝑥4


subject to: 97𝑥1 +  32𝑥2 +  25𝑥3   +  20𝑥4   ≤  139
𝑥𝑖 ∈ {0,  1}, ∀ 𝑖 =  1,  2,  3,  4.

4. Solve the following problem using the branch-and-bound algorithm.

Maximize 10𝑥1 +  2𝑥2 +  7𝑥3   +  5𝑥4


subject to: 𝑥1 +   𝑥2 +   𝑥3         ≤  2
𝑥1 +   𝑥2 +      + 𝑥4    ≤  2
𝑥1 +      + 𝑥3   +   𝑥4    ≤  2
𝑥𝑖 ∈ {0,  1}, ∀ 𝑖 =  1,  2,  3,  4.

Remark: Are Problems 3 and 4 related? If so, how?

5. Solve the following problem using the branch-and-bound algorithm, where the LP relaxations
need to be solved graphically, clearly indicating the feasible regions for each node.

Minimize 50𝑥1 +  100𝑥2


subject to: 7𝑥1 +  2𝑥2   ≥  28
2𝑥1 +  12𝑥2   ≥  24
(𝑥1 , 𝑥2 )   ≥  0, and integer.
6. Consider the minimal cost network flow problem, where the supply/demand at each node are
as specified, and each arc (𝑖,  𝑗) is labeled with the associated 𝑐𝑖𝑗 . Solve the corresponding
minimum cost network flow problem beginning with (𝑥12 ,  𝑥23 ,  𝑥24 ) in the starting basis.
Verify the optimal solution obtained using GAMS.

b1 = 5 b2 = 2
3
1 2

1 5 6

4 3
2
b4 = -3 b3 = -4

7. A machine shop possesses six different drilling machines. On a certain day, five jobs arrive
that need drilling. The number of hours required to perform each job on each machine are
given below. Using the Hungarian algorithm (tableau format), determine the best way to
uniquely assign jobs to machines in order to minimize the total drilling time. List two optimal
solutions. (Hint: Number of jobs and machines must be equal in an assignment problem.)

Job
A B C D E
Machine
1 5 7 6 4 9
2 8 10 3 4 7
3 6 11 5 4 7
4 5 8 7 3 9
5 3 6 4 2 7
6 3 7 5 3 7
8. An intelligence service has 12 agents in an enemy country. In the graph below, each node i
represents an agent, and an arc (𝑖,  𝑗) implies that i and j can communicate with each other.
If i and j communicate, the message passed between them might be intercepted with a
probability 𝑝𝑖𝑗 , as indicated on the graph. The group leader wants to transmit a confidential
message to all the agents while minimizing the total probability that the message is
intercepted. Explain how this problem can be solved as a minimal weight spanning tree
problem and determine the optimal transmission paths and the resulting probability of
interception. Give the corresponding spanning tree graph.

9. Consider the knapsack problem where a thief can potentially steal from a set of n items by
putting them into his knapsack. Each item is associated with its weight 𝑤𝑖 and value 𝑣𝑖 .
However, the knapsack has a maximum weight of W, which is smaller than the total weight
of all items. Therefore, he would like to select a subset of items with the maximum value while
not exceeding the weight capacity of his knapsack. Set-up the network to solve this problem
as a shortest or longest path problem, using the following example with three items and a
knapsack weight capacity of four.

Item Weight Value


A 2 30
B 1 10
C 3 20

S-ar putea să vă placă și