MINIMUM POWER BROADCAST IN WIRELESS

MINIMUM POWER
BROADCAST IN
WIRELESS NETWORKS
Arindam K. Das
CIA Lab
University of Washington
Seattle, WA
MINIMUM POWER
BROADCAST IN
WIRELESS NETWORKS
with
Robert J. Marks II & M.A. El-Sharkawi
(UW CIA)
Payman Arabshahi & Andrew Gray
(JPL/NASA)
Problem Statement
For a designated host and a
broadcast application, find the
connection tree which requires
minimum overall transmission
power.
Example : Minimum Power
Broadcast
Broadcast tree : A  B, C  D
F
C
E
A
D
B
Assumptions (1)
• We assume that there is a fixed source
node which wants to communicate with all
the other nodes in the wireless network
(broadcast).
• All nodes have omni-directional antennas.
• Power is expended for signal
transmission only. No power expenditure
for signal reception or processing.
Assumptions (2)
• The transmitter power is modeled as the
‘’ power of its distance from the receiver
(2    4).
PT  r
Proposed Approach
• We propose a GA based approach for
solving the minimum power broadcast
problem.
• Key question: Encoding of
chromosomes
Some Definitions
• Power matrix, P: The (i,j)th element of the
power matrix is defined as
Pij = rij
where rij is the Euclidean distance between
nodes i and j.
• Cut vector, P: The cut vector, referenced
to P, is an N-element integer vector. It
indicates the location of an element on each
row of the power matrix.
Examples
P
P = [7 2 3 4 3 5 6]
Some Definitions
• Threshold vector, t : An N-element vector
of the elements of P specified by the cut
vector. Represents power settings of the
individual nodes.
• Cost of a cut, c(P) : Sum of the elements
of the threshold vector.
Examples
P
P = [7 2 3 4 3 5 6]
t = [8 0 0 0 2 0 0]
Some Definitions
• Transfer matrix, H: The transfer matrix is
computed by thresholding the power matrix
as follows:
1, if Pij  ti
H ij  
0, otherwise
• Viability of a cut vector: A cut is viable if it
allows all destination nodes to be reached.
Otherwise, it is non-viable. A viable cut
vector has an associated connection tree.
Examples
P
P = [7 2 3 4 3 5 6]
t = [8 0 0 0 2 0 0]
Solution Approach
“As Implemented”
• GA based
• Chromosome encoding : cut vectors, P.
• Crossover : random 1-point crossover,
subject to a certain crossover probability.
• Parent selection : roulette wheel
• Fitness function : c(P)
• Mutation : none
• Elitism : yes
Viability of the Children
• Randomly generated cut vectors need not
be viable  the children created after
crossover and mutation need not correspond
to viable connection trees.
• Use the Viability Lemma to determine the
viability of a child.
- If viable, accept it.
- If not, reject it, or, apply a repair operator.
Viability of the Children
A Repair Strategy
• Suppose a node (say n) is not reached by
a cut.
• Identify the node closest to n (say m).
• Augment the power level of m so that node
n is reached and modify the mth element of
the cut accordingly.
Viability Lemma (1)
• Notation
k = iteration index
 = N-element binary node coverage vector
• Nodes which are reached are tagged by
a ‘1’ in the coverage vector. Nodes not
reached are tagged by a ‘0’.
Viability Lemma (2)
• Initialize (0) = [0 0 .. 1.. 0 0].
All elements, except that corresponding to
the source, are set to 0.
•   logical product of two matrices
(multiplications replaced by AND’s and
additions replaced by OR’s).
• Apply the iteration
(k+1) = HT  (k)
Viability Lemma (3)
• Necessary and sufficient condition for
a cut to be viable (assuming broadcast
application)

(N -1) = 1
• The iteration process terminates if
(K)

= 1, K  N  1
Generating the Initial Gene Pool
• The initial gene pool is generated using an
iterative, random node selection method
(the Stochastic Tree Generation
algorithm).
• Rules:
– First transmission must be from source.
– A node can transmit only once.
– A transmitting node, in general, can opt to be a
leaf, if choosing so does not render the tree
nonviable.
Generating the Initial Gene Pool
Example
Iteration 1
• Assume node 1 is the source.
1
Possible Transmitting
Nodes
2, 3, 4, 5, 6
Possible Destination
Nodes
• Transmitting node = 1
• Randomly chosen destination node = 3
Generating the Initial Gene Pool
Example
Iteration 2
• Assume 1  3 also reaches node 4.
3, 4
Possible Transmitting
Nodes
2, 3, 5, 6
Possible Destination
Nodes
• Randomly chosen transmitting node = 3
• Randomly chosen destination node = 3
Generating the Initial Gene Pool
Example
Iteration 3
• Assume 4  6 also reaches node 5.
4
Possible Transmitting
Nodes
[ …], 5, 6
Possible Destination
Nodes
• Randomly chosen transmitting node = 4
• Randomly chosen destination node = 6
Generating the Initial Gene Pool
Example
• Converting the transmission sequence to a
cut vector, P.
1
2
3
4
3
13
2
33
3
46
6
5
5
6
6
Simulation Results
• Simulations on 50 randomly generated 25-node
and 50-node networks show an improvement of
approximately 10% and 13% over the solutions
generated using the Broadcast Incremental
Power algorithm proposed by Wieselthier et al.
• Simulations were conducted using 100
chromosomes and 50 evolutions.
Summary
• Discussed a GA based search method for
solving the minimum power broadcast
problem in wireless networks.
• Discussed the Stochastic Tree Generation
algorithm for generating the initial population.
Solutions from other heuristics can be included
in the initial population.
• Discussed the computationally simple
Viability Lemma for determining the viability
of the children.