| dc.description.abstract | Opportunistic selection reaps the benefits of multiuser diversity in many wireless
applications. For example, it increases throughput of cellular networks, lifetime
of wireless sensor networks, and improves diversity in cooperative relay systems.
In it, the best node is selected for data transmission. The ability of a node to
improve the system performance is quantified by a real-valued metric. The best
node is the one with the highest metric.
Since the nodes are geographically separated, no node in the network knows
beforehand who the best node is. Hence, distributed selection schemes are required.
We investigate the role of power control in distributed selection schemes.
Power control allows the nodes to choose their target receive power from a set
of pre-specified power levels. These levels are set such that a node can be selected
even if some other nodes transmit with it but do so with lower power levels.
Without power control, this would have resulted in a wasteful collision.
First, we propose a new timer-based selection scheme with power control. In
it, each node sets its timer and its target receive power level as a function of its
metric. We develop several structural insights, about the optimal metric-to-timerand-
power mapping, which maximizes the probability of selecting the best node.
These significantly reduce the computational complexity of finding an optimal
mapping and lead to valuable asymptotic insights. We show that the proposed
scheme significantly outperforms the conventional timer-based selection scheme.
Performance improvements as large as 28.2% are achieved even with only two
target receive power levels.
We also analyze the performance of the proposed scheme with imperfect power
control, which is inevitable in practice. It randomly changes a node’s receive power
and, thus, its signal-to-interference-and-noise ratio. Consequently, it affects the
probability of selection of the best node and of no node getting selected. It can
even lead to a sub-optimal node getting selected. We derive the probability of
each of these effects. We observe that by increasing the lowest target receive
power level enough, the reduction in probability of selection of the best node can
be restricted to be within 9%.
Lastly, we characterize the effect of imperfect power control on an alternate
scheme known as the splitting-based selection scheme with power control. This
scheme maintains a set of thresholds for each slot, which determine the nodes that
transmit in that slot and their target receive power levels. The outcome of a slot
is used to update the thresholds for the next slot. In case of a collision, the set of
colliding nodes is split in two smaller subsets, the nodes in one of which transmit
in the next slot. We analyze the average number of slots required for selection
with perfect and imperfect power control. We also compute the probability that
the best node or a sub-optimal node gets selected in each slot, and the probability
that no node ever gets selected due to imperfect power control. The implications
of imperfect power control on system throughput are also investigated. We show
that imperfect power control reduces the system throughput. We also evaluate
the effect of peak power constraint on the performance of the timer-based and
splitting-based selection schemes with power control. We show that it can reduce
the probability of selection of the two schemes by as much as 19.1% and 9.7%,
respectively, when the metric and channel gain are uncorrelated depending on the
system parameters. | en_US |
