BDC5101: Deterministic Operations Research Models Semester I, 2019/2020, NUS

Assignment 3: Due on Nov.1, 2019

1. (10’) Process Flexibility: Consider the case of the Ford Motor Company discussed in class. Suppose
the company owns 6 production plants and is selling 6 models. The capacity at each plant is 100
per week, and the demands for each model are assumed to be independent and following N (100, 302 )
truncated above zero. Recall that a dedicated design results in an average sales of around 525 while a
full flexible design results in an average sales of around 570. Evaluate the performance of the following
two chain designs.

(a) Open Chain Design (b) Long Chain Design

2. (10’) Multi-Location Newsvendor: Consider a fashion retailer in Singapore that manages three
retail stores at Jurong West, Orchard and Harbour Front respectively. The retailer faces the prob-
lem of how to manage the inventories of its best-selling product at these three locations for the
new selling season. The product of concern has a retail price of S$100 at all three locations and
costs S$50 per unit to procure from the manufacturer. The demands for the product at location
i, i ∈ {Jurong West, Orchard, Harbour Front} are assumed to be normal truncated above zero, i.e.,
max{N (µi , σi2 ), 0} and the corresponding parameters, estimated from historical data, are summarized
in the first two rows in the table below. The demands across different locations are assumed to be
Table 3.1: Estimation of Demand Parameters and Inventory Position

Jurong West Orchard Harbour Front

µi 300 500 500
σi 20 20 40
Inventory 300 500 500

independent. The management team decides to procure in total 1300 products from the manufacturer

3-1
Assignment 3: Due on Nov.1, 2019 3-2

and put 300, 500 and 500 units at Jurong West, Orchard and Harbour Front respectively (see last
row in Table 3.1). That is, the inventory at a particular location is matched with the mean demand
(before truncation). No inventory can be replenished from the manufacturer during the selling season
and unsold products are worthless after the selling season.
(a) Compute the total expected profit aggregated over the three locations (the model of process
flexibility example discussed in class can be used to compute this with minor modification).
(b) What is the total expected profit if there is no demand uncertainty at all (i.e., σi = 0)?
(c) To deal with the demand uncertainty, the management team is considering employing a logistic
service company to help transship its products from one store to another. For example, after the
selling season begins, the demand is realized to be 250, 550 and 450 units at Jurong West, Orchard
and Harbour Front respectively. In this case, Orchard is running out of inventory but both Jurong
West and Harbour Front have some leftover. The logistic company can then transship either 50
units of leftover at Jurong West or Harbour Front to satisfy the remaining demand at Orchard.
The logistic company charges a fixed transaction cost of S$200 (which is incurred even if nothing
is shipped) and a unit transshipment cost depending on the origin and destination summarized
in the table below. That is, if 50 units of product were to be shipped from Harbour Front to
Orchard, then the total cost to the retailer is S$200 + 7 × 50. Should the retailer adopt the service

Table 3.2: Transshipment Cost (S$/unit)

Jurong West Orchard Harbour Front

Jurong West 0 22 19
Orchard 22 0 7
Harbour Front 19 7 0

provided by the logistic company (in simulation, you should check the robustness of your answer
by looking at larger sample size and running simulations multiple times)?

3. (10’) Consider a transportation problem with two supply nodes and two demand nodes. The supply
nodes have a and 100 − a supplies respectively with 0 ≤ a ≤ 100. The demand nodes require 70 and
30 respectively. The figure below illustrates the graph of the transportation problem. Let cij be the
unit transportation cost from supply i to demand j, 1 ≤ i, j ≤ 2.

Supply Demand

a 1 1 70

100- a 2 2 30

(a) Write down a linear programming formulation of the above problem and present your A matrix.
(b) Consider a spanning tree of the above graph generated by deleting the arc (1, 1) (or forcing
x11 = 0). Find the flow solution that satisfies all the flow-balancing constraints (i.e., the solution
to Ax = b).
Assignment 3: Due on Nov.1, 2019 3-3

(c) Provide the range of values on the supply a such that the flow solution you found in (b) is a feasible
solution to the transportation problem.

4. (10’) Let x1 , x2 , . . . , xn be n binary variables and let z be another binary variable. Formulate the
following logical statements as linear constraints.
(a) z takes the value of 1 if at least one of the xi ’s takes the value of 1.
(b) If all the xi ’s take the value of zero, then z takes a value of 0.
(c) z takes the value of 1 if all the xi ’s take the value of 1.
(d) If any xi takes the value of zero, then z takes a value of 0.
(e) If x1 + x2 + · · · + xn ≥ 5, then z takes a value of 1.

5. (10’) TSP with Time Window: Consider the problem faced by an online food delivery company
(e.g., Deliveroo). The following map illustrates the location for picking up the food (location 1) and
the locations of each orders (locations 2-6) around the area of Vivo City in Singapore. The data
for coordinates of each location is presented in Table 3.3. The last column in Table 3.3 provides

Table 3.3: Data on locations

Latitude Longitude Maximum Waiting Time (in minute)

1 1.264306 103.822222 -
2 1.265270 103.821762 5
3 1.264181 103.820403 10
4 1.265067 103.818683 15
5 1.265655 103.818049 13
6 1.264998 103.817868 14

information on the maximum number of minutes a customer at a particular location is willing to

wait before filing a complaint. With the data on latitude and longitude, one can use Google map to
Assignment 3: Due on Nov.1, 2019 3-4

Table 3.4: Pairwise Walking Time

1 2 3 4 5 6
1 - 3 3 10 9 10
2 3 - 3 7 6 7
3 3 3 - 7 6 7
4 10 7 7 - 1 2
5 9 6 6 1 - 1
6 10 7 7 2 1 -

compute the pairwise walking time (in minute) summarized in the following table. Suppose now the
delivery man is already at location 1 to pick up the food. The delivery man needs to deliver the food
to customers at all the other locations and then return back to location 1. Assuming the time for
consumers to pick up their delivery is zero, the objective of the delivery problem is to minimize the
total traveling time (or waiting time of the consumers).
(a) Without considering the maximum waiting time of the consumer, what is the minimum traveling
time and route to traverse the locations of all customers and returning to location 1? (the route
may not be unique)
(b) To ensure a good service quality, the delivery company commits to the delivery before consumers’
maximum waiting time. Can you still find a feasible route in this case? If yes, what is the
minimum traveling time and route?