Digital Signal Processing Reference
In-Depth Information
2D contour curves are in reality not constructed, but nodes locally make a binary
classification as being interior or not. The curves are shown only for illustration
purposes.
The images in row A show the secondary contour resulting from the initial power
estimate based on the distance between the secondary sender and the primary con-
tour. An optimistic pathloss model with
α
2 was used for the computation of the
secondary transmitter's power as function of the measured distance from the first
transmitter. Since actual propagation is by definition worse than the most optimistic
condition, the coverage of the secondary transmitter is typically very small using
this power, as can be seen in Fig.
4.7
,rowA.
The algorithm however iteratively estimates a more realistic pathloss model for
the secondary transmitter, and as a result achieves a much more realistic power ad-
justment already after one iteration. As can be seen in Fig.
4.7
, the resulting coverage
of the secondary transmitter can be increased considerably in each of the 3 scenar-
ios, without causing interference (i.e., overlap of the contours) to the first transmit-
ter. Each power adjustment step requires estimation of the secondary transmitter's
contour at the cost of
N
transmissions. Next, each node inside the contour requires
propagating its distance to the first contour and its estimated pathloss, at the cost
of
N
interior
<N
transmissions. As can be seen in the final images C, a much larger
area is covered by the secondary transmitter as compared to the initial configuration
shown in row A.
The algorithm not only results in improved spatial reuse gain, but also relaxes the
opportunistic transmitters' sensitivity requirements. Indeed, traditional approaches
would require each possible transmitter to be able to sense the first transmitter. For
a given target power of the second transmitter, this results in very high sensitivity
requirements. In this distributed run time approach, these requirements are relaxed,
and the first transmitter's propagation contour is estimated close to the first transmit-
ter and then efficiently propagated. Both this estimation and propagation are cheap
and require only
N
transmissions. Typically, the algorithm converges in two to three
iterations.
Due to the approximative inside/outside classification, we can not guarantee that
the real final contours do not overlap and little interference might occur. As can
be seen on the images, only in example 2 the final contours overlap, though for a
negligible amount. However, it would be interesting future research to quantify the
amount of possible overlap.
=
4.4 Conclusions
In this chapter, focus was on the monitoring challenge for smart opportunistic ra-
dios. A new approach for iterative power control where to goal is to maximize
spatial reuse while avoiding unwanted interference to existing networks was pro-
posed. First, the propagation contours of the existing networks are estimated and
efficiently propagated. Next, the secondary transmitter's power control is iteratively
