You are on page 1of 4

ENGG1400 – Workshop for Week 6

PART I
NETWORK DESIGN: THE MINIMUM-COST FLOW PROBLEM

Network design is an important infrastructure problem which can provide a solution to several
types of problems, such as:

• How to supply a maximum amount of flow through a network?


• How to improve traffic network performance?
• How to optimize the location of transit infrastructure?
• How to minimize the operations cost in a telecommunication network?
In this workshop, we will focus on a fundamental network design problem which is known as the
Minimum-cost Flow Problem. This problem can be defined as follows: given a set of nodes and a
set of capacitated arcs that form a network, what is the cheapest way to satisfy a flow demand
between an origin node and a destination node? In the Minimum-cost Flow Problem (MinCost),
each arc has a weight which represents the cost of propagating a unit flow from its tail node and to
its head node. The decisions of interest in MinCost is the amount of flow that should be sent
through each arc of network in order to satisfy the flow demand and minimize the total cost.
Problem: Information flow in telecommunication network
Consider a bi-directed telecommunication network presented in figure 1. There are several routes
that can be used to pass the information from a source node to a destination node. Find the best way
to pass the information so as to minimize the total operations cost.

Fig. 1 – An example of network for the Minimum-cost Flow Problem


Figure 1 depicts the telecommunication network, the associated cost per unit of information
transferred and the capacity of each link in the network. The links are bi-directed and have the same
capacity and operating cost on both the directions.
Find the best way to transfer 30 units of information from node A to node E. How many units of
information should be passed on each route of the network to minimize the total cost?
The AMPL files relative to this network design problem are contained in the compressed file
Week_6.zip which can be downloaded from Moodle. After extracting this compressed file, you
should have the following files in the folder ...\ENG1400\Week_6:
w6_MinCost.dat
w6_MinCost.mod
w6_MinCost.run
Similarly to the other workshops, we will examine different scenarios where the capacity of the arcs
of the network and the demand change. Hence eventually new information relevant to this
optimization problem will appear and you will often be modifying your .dat and .mod files so as to
integrate these changes in the mathematical formulation of the problem. Keep in mind that you can
always re-download the original files from Moodle if you need to start over.
The model for the MinCost can be mathematically stated as:

𝑚𝑖𝑛 ∑ 𝑤𝑖𝑗 𝑥𝑖𝑗 (1)


(𝑖,𝑗)∈𝐴
Subject to:
𝑆𝑖 + ∑ 𝑥𝑗𝑖 = 𝐷𝑖 + ∑ 𝑥𝑖𝑗 ∀𝑖 ∈ 𝑁 (2)
(𝑗,𝑖)∈𝐴 (𝑖,𝑗)∈𝐴
𝑥𝑖𝑗 ≤ 𝑐𝑖𝑗 ∀(𝑖, 𝑗) ∈ 𝐴 (3)
𝑥𝑖𝑗 ≥ 0 ∀(𝑖, 𝑗) ∈ 𝐴 (4)
N is the set of nodes, and A is the set of possible routes between nodes. xij represents the flow on
arc (i, j) ∈ A, wij is the weight of arc (i, j) and cij is the capacity of arc (i, j).

Si represents the supply at node i, Di represents the demand at node i. Note that some nodes may
have a null supply and demand (e.g. a transit node), some nodes may only be supplier, some may
only be demand nodes and some may be both supply and demand nodes.
This model is given in the file w6_MinCost.mod and uses the same type of mathematical terms that
have been used in the workshops so far (sets, parameters, variables, objective function and
constraints) and can be solved using the scrip w6_MinCost.run.

QUESTIONS

Question 1:
What is the optimal flow on each arc of the network and what is the total operations cost? Are some
nodes never used to pass the information in this case?
Question 2: Assume now that the supply of source node is 40 units and it transfer information to
node D in addition to node E. The demand of node D is 10 units.
How can we adapt the AMPL model and data files to incorporate this new business model? What is
the new optimal flow on each arc of the network and what is the new total operations cost? Are
some nodes never used to pass the information in this case?
PART II

NETWORK DESIGN: THE CAPACITY ALLOCATION PROBLEM


I. Problem Description
This workshop is an extension of the workshop for Week 6 on Network Design. In Week 6, we
have studied the Minimum Cost flow problem that sought the best assignment of commodities to
links of the network in order to minimize transportation costs. This week, we focus on the impact of
allocating capacity improvements in the network.

We build on the same illustration that was used in Week 6: Information flow in
telecommunication network. Hence, we assume that the reader is familiar with the problem
context and its mathematical representation.
For the remainder of this workshop, we assume that the company has the possibility to increase the
capacity of some of the arcs of the network and we examine what is the best strategy to allocate
these capacity improvements.

The AMPL files related to this network design problem are contained in the compressed file
Week_6.zip which can be downloaded from Moodle. After extracting this compressed file, you
should have the following files in the folder ...\ENG1400\Week_6:
w6_Network_Design.dat
w6_Network_Design.mod
w6_Network_Design.run
w6_Network_Design_loop.run

Information flow in telecommunication network:

• Bi-directed telecommunication network


• Cost per unit of information transmitted and capacity of each link are given in the network
• The company can increase the capacity of at most B links.
• Additional capacity does not affect total cost. This is a separated deal.
• Capacity of each arc can be added either 5 units or nothing.
• Find the best way to transmit 30 units of information from node A to node E.
• How many units of information should be passed on each route of the network to minimize
the total cost?

Your demonstrators will now help you derive the mathematical representation for this
problem. You will then use the formulation to answer the questions below.

QUESTIONS
Question 1:
1a: In the case where 1 capacity improvement is available, what is the total operations cost
induced? On which arc is the capacity improvement allocated. Show the optimal solution on a
diagram.

1b: In the case where no capacity improvements are available, what is the total operations cost
induced? Show the optimal solution on a diagram. Could we have used a simpler model to answer
this question?

Question 2:
2 Assume that the company has decided to send 40 units of information from A to E. In the case
where 10 capacity improvements are available, what is the lowest number of capacity
improvements that are necessary to provide the minimal operations cost?

You might also like