You are on page 1of 3

University of Tunis

Tunis Business School Spring 2021

BA 310: Operations Research


Midterm Examination Solution

Question 1 (24 points)


Consider the following constraints in a mathematical programming problem:
3x1 + 2x2 -x3 ≤ 9 (1)
2x1 + x2 +x3 ≤ 12 (2)
x1 - 2x2 +3x3 ≥ 4 (3)
Formulate each of the following statements to be added to the problem:
1. Exactly (1) or (2) must occur. (5 points)
3x1 + 2x2 -x3 ≤ 9+MY1
2x1 + x2 +x3 ≤ 12+ M(1-Y1)
Y1 = 0, 1
2. At least (1) or (2) must occur. (5 points)
3x1 + 2x2 -x3 ≤ 9+MY1
2x1 + x2 +x3 ≤ 12+ MY2
Y1+Y2 ≤1
Y1, Y2 = 0, 1
3. If (1) occurs, then (2) must occur. (5 points)
3x1 + 2x2 -x3 ≤ 9+MY1
2x1 + x2 +x3 ≤ 12+ MY2
Y2 ≤ Y1
Y1, Y2 = 0, 1
4. If (1) or (2) occurs, then (3) must occur. (9 points)
3x1 + 2x2 -x3 ≤ 9+MY1
2x1 + x2 +x3 ≤ 12+ MY2
x1 - 2x2 +3x3 ≥ 4-MY3
2Y3 ≤ Y1+Y2
Y1, Y2, Y3 = 0, 1

Question 2 (26 points)


Suppose that the Branch & Bound method is applied to a particular IP problem and
yields the following situation:

L0 Z0 =82, x =(4, 1.5)


x21
x22

L1
Z1=78,x=(6,1) L2 Z2=78.5, x=(3.5,2)

x13 x14

Z3=74 L3
L4
x=(3, 3)

Infeasible
1. Is the IP feasible? Justify Yes. (6, 1) is a feasible solution. (4 points)
2. Is it a maximization or a minimization problem? As Z1< Z0, then it is a maximization
IP (4 points)
3. Provide an upper bound on the IP objective value. 82 as given by Z0 is an upper
bound (4 points)
4. Provide a lower bound on the IP objective value. Z1 provides a lower bound of
78. (4 points)
5. Is it possible to conclude about a solution or more iterations are still needed? L1 in fact
provides an optimal solution given that all subproblems are fathomed. (5 points)
6. Was it possible to conclude earlier? If the objective function coefficients are integer,
then we may conclude at the level of L2 given that the optimal objective value must be
integer less than 78.5 and L1 has already provided a LB of 78, then L1 provides the
optimal solution in that case. Else, it is not possible to conclude earlier. (5 points)

Question 3 (50 points)

Solve the shortest path problem from node A to node J using backward dynamic
programming.

B E

H
A C F
J

I
D G

*
The distance from Node A to Node B is 5 denoted as d(A, B) =5.
Similarly, d(A, C) =6, d(A, D) =7, d(B, E) =8, d(B, F) =10, d(B, G) =5, d(C, F) =6,
d(D, E) =9, d(D, F) =4, d(D, G) =1, d(E, H) =5, d(E, I) =6, d(F, H) =9, d(F, I) =5,
d(G, H) =7, d(G, I) =3, d(H, J) =2, d(I, J) =4

1. Identify the DP stages of the problem. 4 stages: from A; stage 1 to B, C, or D; then


stage 2 to E,F, or G; then stage 3 to H or I; and finally stage 4 to J. (6 points)
2. Determine the decision variables. Let xn be the immediate destination on stage n
with x4 = J and x1=A. Let s be the current state (node). Let fn(s, xn) be the shortest
remaining distance from node s at stage n to node J. Let xn* minimize fn(s, xn). Define
f*n(s) = 𝑚𝑖𝑛𝑥𝑛 𝑓𝑛 (s, xn)= fn(s, xn*) (15 points)
3. Determine the recursive formula. fn(s, xn)=d(s, xn)+ f*n+1(xn) (5 points)
4. Perform the adequate iterations leading to the shortest path.
n=4
s f*4(s) x4*

H 2 J

I 4 J

(4 points)

n=3
x3 f3(s, x3)=d(s, x3)+f*4(x3)
s H I f*3(s) x3*
E 5+2=7 6+4=10 7 H
F 9+2=11 5+4=9 9 I
G 7+2=9 3+4=7 7 I
(6 points)

n=2
x2 f2(s, x2)=d(s, x2)+f*3(x2)
s E F G f*2(s) x2*
B 8+7=15 10+11=21 5+7=12 12 G
C - 6+11=17 - 17 F
D 9+7=16 4+9=13 1+7=8 8 G
(6 points)
n=1
x1 f1(s, x1)=d(s, x1)+f*2(x1)
s B C D f*1(s) x1*
A 5+12=17 6+17=23 7+8=15 15 D
(4 points)
It follows that the optimal path is A-D-G-I-J with total distance of 15. (4 points)

You might also like