INDU 6141 Winter 2024

Final Exam

Submit by May 1, 2024

Instructions
1. This nal is to be completed and submitted individually.

2. No form of plagiarism will be tolerated (yes, including LLMs).

3. To submit, please drop your answers in my mailbox, next to the kitchenette, EV 4th oor.

Exercise #1

Note: unless you code a parametrized version of these methods, it will be impossible to do by hand an
exhaustive analysis for all values of α. Please solve for α ∈ {0.0, 0.2, 0.4, 0.6, 0.8, 1.0} to drive your analysis.

Solution
The score s(i) of a supplier is computed as

AW SM (i) = αpi + (1 − α)ri

   
pi ri
RW SM (i) = α + (1 − α) .
pw rw

1
 In this case pw = 3.8, rw = 1.2 denote the worst indicators for the price and reliability among all 6 suppliers.
For α = 0.4 we obtain that the best absolute score AW SM (i) is obtained by supplier 5 (AW SM (5) = 3.04)
while the best relative score is obtained by supplier 6, with RW SM (6) = 1.355. For the dierent values of
α ∈ {0, 0.2, 0.4, 0.6, 0.8, 1} we obtain the following best scores:

α supplier score
0.0 4 1.90
0.2 5/6 2.32
0.4 5 3.04
0.6 3 3.78
0.8 3 4.64
1.0 3 5.50

Table 1: Best absolute scores

α supplier score
0.0 4 1.58
0.2 4 1.47
0.4 6 1.36
0.6 5 1.35
0.8 5 1.36
1.0 3 1.45

Table 2: Best relative scores

We observe that the choice of the supplier is highly dependent on the type of score function (absolute vs
relative) and the parameter α. We also observe that suppliers 1, 2 do not make the cut in any combination.

Exercise #2

2
Solution
The capacity of the storage area is equal to 12 × 12 × 6 = 864 pallets.
To compute the throughput, let us rst compute Lx , Ly using the data provided:

Lx = 12 × 1.25m + 6 × 3m = 33m
Ly = 3m + 12 × 0.91m = 13.92.

In addition, a forklift will have to ascend/descend for 6 piled racks ((6 − 1) × 2.25)/2 = 5.625m on average.
A forklift retrieving a pallet will have to perform the following operations (in order):

# Operation Avg time (s)

1 Start empty from the loading/unloading zone 
2 Go left empty on the x-axis (33/2)/3.75 = 4.4
3 Go up on the y -axis empty (13.92/2)/3.75 = 1.856
4 Go up on the z -axis empty 5.625/0.5 = 11.25
5 Pickup a pallet 70
6 Go down on the z -axis loaded 5.625/0.45 = 12.5
7 Go down on the y -axis loaded (13.92/2)/3.15 = 2.21
8 Go right on the x-axis loaded (33/2)/3.15 = 5.24
9 Unload the pallet in the loading/unloading zone (included in 5)
107.5

A forklift storing a pallet will have to perform the following operations (in order):

# Operation Avg time (s)

1 Pickup a pallet from the loading/unloading zone 70
2 Go left loaded on the x-axis (33/2)/3.15 = 5.24
3 Go up on the y -axis loaded (13.92/2)/3.15 = 2.21
4 Go up on the z -axis loaded 5.625/0.25 = 22.5
5 Drop a pallet (included in 1)
6 Go down on the z -axis unloaded 5.625/0.5 = 11.25
7 Go down on the y -axis unloaded (13.92/2)/3.75 = 1.856
8 Go right on the x-axis unloaded (33/2)/3.75 = 4.4
9 Arrive empty to the loading/unloading zone 
117.5

The average operation time for storing/retrieval operations is equal to 107.5 + 117.7 = 112.5 seconds, or
0.03125 hours. The hourly throughput of a single forklift is equal to 1/03125 = 32 pallets per hour. In a
typical day (8 hours of operation) it will be of 256 pallets/day. Three forklifts will triple this throughput, to
768 pallets/day. If the eciency of a forklift decreases to η = 0.7, we will need ⌈3/0.7⌉ = ⌈4.28⌉ = 5 forklifts to
keep the same or a higher throughput.

3
Exercise #3

Solution
If nx , ny represent the number of storage locations on the x- and the y -axis respectively, we have that
Lx = 2 × 1.5 + 1.2nx
Ly = (2.5/2 + 0.803)ny
A picker entering/leaving by zone 1 will have to travel Lx + Ly meters on average, while a picker enter-
ing/leaving from zone 2 will travel Lx + Ly /2 on average. If 0 ≤ p ≤ 1 represents the proportion of pickers
entering by zone 1, and (1 − p) those who enter from zone 2, the optimal storage area is giving by the following
optimization model that minimizes the average traveling distance:
Minimize Z = p(Lx + Ly ) + (1 − p)(Lx + Ly /2)
subject to
25nx ny ≥ 60, 000
nx , ny ∈ Z+
ny even.
Note that the above model can be reduced to
Minimize Z = 3 + 1.2nx + 1.0265(1 + p)ny
subject to
nx ny ≥ 2400
nx , ny ∈ Z+
ny even.
If we relax the integrality constraints and assume that the capacity constraint will be binding, we can use
the substitution nx = 2400/ny , which leads to the model
Minimize Z = 3 + 2, 880/ny + 1.0265(1 + p)ny
subject to ny ≥ 0.

4
 q
An optimal solution of this relaxation is given by n∗y = 1.0265(1+p)
2880
, n∗x = 2400/n∗y . A feasible solution to
the integer problem can be obtained by rounding to the next integer in the case of n∗x and to the next even
integer in the case of n∗y .

Exercise #4
Consider the traveling salesman problem including one depot (node 1) and 9 customers (nodes 2. . . 10) dened
by the following distance matrix:

2 3 4 5 6 7 8 9 10
1 57 47 72 80 66 96 60 60 23
2  17 47 37 39 44 42 16 46
3   36 53 51 50 52 30 43
4    82 85 45 87 62 76
5     20 60 30 24 61
6      73 10 24 44
7       80 52 89
8        28 37
9         44

Execute a 2-opt heuristic starting from the sequence r = (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1). Clearly indicate at
every iteration: i) the move performed; ii) the cost reduction achieved.

Solution
In the 2-opt heuristic we iteratively replace two non-consecutive edges with two other edges with smaller total
cost.

Iteration Path Remove Add ∆

57 17 36 82 20 73 80 28 44 23 82 80 45 30
1 1 −→ 2 −→ 3 −→ 4 −→ 5 −→ 6 −→ 7 −→ 8 −→ 9 −→ 10 −→ 1 4 −→ 5, 7 −→ 8 4 −→ 7, 5 −→ 8 -87
57 17 36 45 73 20 30 28 44 23 73 44 52 44
2 1 −→ 2 −→ 3 −→ 4 −→ 7 −→ 6 −→ 5 −→ 8 −→ 9 −→ 10 −→ 1 7 −→ 6, 9 −→ 10 7 −→ 9, 6 −→ 10 -21
57 17 36 45 52 28 30 20 44 23 30 44 10 61
3 1 −→ 2 −→ 3 −→ 4 −→ 7 −→ 9 −→ 8 −→ 5 −→ 6 −→ 10 −→ 1 8 −→ 5, 6 −→ 10 8 −→ 6, 5 −→ 10 -3
57 17 36 45 52 28 10 20 61 23 52 61 60 44
4 1 −→ 2 −→ 3 −→ 4 −→ 7 −→ 9 −→ 8 −→ 6 −→ 5 −→ 10 −→ 1 7 −→ 9, 5 −→ 10 7 −→ 5, 9 −→ 10 -9
57 17 36 45 60 20 10 28 44 23
5 1 −→ 2 −→ 3 −→ 4 −→ 7 −→ 5 −→ 6 −→ 8 −→ 9 −→ 10 −→ 1 

Note that there may be other operations leading to decreases in the cost, which would lead to other solutions.
Any of those solutions would be alright.