You are on page 1of 3

BİL 512

HW2

1) Relaying systems:

Sink

3C
C C

Relay
4C 4C

Source 1 Source 2

Consider the relaying system illustrated in the figure above. In this system there are two
source nodes who want to transmit data to a sink. They can split their transmission time
between direct link to the sink (with capacity C) and indirect link through the relay. The
capacity of the links between the source nodes and the relay is 4C. However, there is a
bottleneck link capacity of 3C between the relay and the sink. The source nodes control
their routing through a parameter y i which dictates percentage of their time they transmit
to the relay (direct connection to the sink and the connections to relays are orthogonal in
frequency). Eg. if a source node uses direct connection 40% of the time, its utilizes 0.4C
through direct connection. Assume that the relay shuts off when the total incoming
capacity is greater then the bottleneck (3C). Each user tries to maximize their utilities
which correspond to their total outgoing capacity:

1  y i C  y i 4C if y i 4C  y j 4C  3C
U i y i , y j   
1  y i C ow.
I. Plot U i  y i , y j  as a function of y i for fixed y j .

II. Do Nash equilibria exist? Which one is the most fair equilibrium?

III. One of the challenges in dealing with distributed systems is to come up with
distributed algorithms that actually achieve the Nash equilibrium. Such algorithms are
usually in the form of iterative update algorithms in which the players iteratively
modify their strategies in a round robin fashion, considering the opponent’s strategy
from the previous iteration. (See algorithm 1 in “Efficient Power Control via Pricing
in Wireless Data Networks” by Cem Saraydar as an example). Note that the best
response dynamics does not necessarily achieve a Nash equilibrium. Even if it does,
the achieved equilibrium is not necessarily the Pareto dominant one.

Consider the following algorithm (take from the paper mentioned above) and answer
the following questions.

Suppose terminal j=1,2 updates its strategy γj at time instances given by the set
Tj={tj1,tj2,tj3,…} where tjk<tj(k+1) and tj0=0 for all j. Define T={τ1, τ2,…}as the set of
update instances T1UT2U….UTN sorted in the increasing order.

Algorithm:
1) Set the initial strategies to γ1=0 and γ2=0.
2) For all k such that τk ϵ T
i. Given γ1(τk-1) and γ2(τk-1), compute γ*j (Compute the best response of
the user updating its strategy at this time instant, given the opponents
recent strategy)

Questions:
1- Assume that the nodes update their strategies at different time instants (T1 
T2= 0 ). Does the algorithm converge to an equilibrium? If yes, which equilibrium
does it converge to? Comment on the fairness.
2- Assume that the nodes are synchronized, eg. they update their strategies at the
same time instant (t11= t21, t12= t22,….. ). Does this algorithm converge? If yes,
which equilibrium does it converge to? Comment on fairness.

IV. Now assume that when the bottleneck constraint is saturated, the relay forwards at
most 3/2C for each user. Modify the utility function accordingly and answer parts I
and II for the modified case.
2- ) Consider the dynamic game below:

a) List the strategy spaces for both players.


b) Can you use backward induction to find a solution? Why?
c) Find the subgame perfect NE of the game.

3) Consider a duopoly game between firms A and B, in which they both have
the choice of selecting a high or a low price for the good they produce. They make
the choice of price simultaneously every year, and the game is played repeatedly,
year after year. If the firms both select the high price, they both earn $80.000 per
year. If one sets a high price, and the other sets a low price, the low price firm earns
$100.000, and the high price firm earns $30.000. If they both set a low price, they will
both earn $50.000.

a) Construct the payoff matrix of the one shot game and find the Nash eq.
for this case.

b) Assume that the firms decide to play this game for a period of 3 years.
Determine the game tree. What are the total equilibrium payoffs for this
scenario?
c) Assume now that the firms are going to play this game forever, and apply a
discount factor δ = 0.8. If both firms play the “Grim-Trigger” strategy (both price
high, until one of them defects and they both price low for the rest of the game). Will
the firm have incentives to cooperate for this scenario?

You might also like