On node density - outage probability tradeoff in wireless networks

Abstract

Abstract-A statistical model of interference in wireless networks is considered, which is based on the traditional propagation channel model and a Poisson model of random spatial distribution of nodes in 1-D, 2-D and 3-D spaces with both uniform and non-uniform densities. The power of nearest interferer is used as a major performance indicator, instead of a traditionallyused total interference power, since at the low outage region, they have the same statistics so that the former is an accurate approximation of the latter. This simplifies the problem significantly and allows one to develop a unified framework for the outage probability analysis, including the impacts of complete/partial interference cancelation, of different types of fading and of linear filtering, either alone or in combination with each other. When a given number of nearest interferers are completely canceled, the outage probability is shown to scale down exponentially in this number. Three different models of partial cancelation are considered and compared via their outage probabilities. The partial cancelation level required to eliminate the impact of an interferer is quantified. The effect of a broad class of fading processes (including all popular fading models) is included in the analysis in a straightforward way, which can be positive or negative depending on a particular model and propagation/system parameters. The positive effect of linear filtering (e.g. by directional antennas) is quantified via a new statistical selectivity parameter. The analysis results in formulation of a tradeoff relationship between the network density and the outage probability, which is a result of the interplay between random geometry of node locations, the propagation path loss and the distortion effects at the victim receiver.

Performance of Optimum Combining in a Poisson Field of Interferers and Rayleigh Fading Channels

“…In our case, despite the large number of users (the network area is wide), these results cannot be applied. In fact, the spatial separations of users ensures that the global interference is only influenced by a small number of them [8], [15], [16]. In addition, the goal of our work is to show how using a small number of receive antennas can improve the SINR.…”

“…In [8], [15] and [16], approximations and bounds on the outage probability are given for the successive interference cancellation receiver. In the context of smart antennas, [17] gives lower and upper bounds on the outage probability for sectorized antenna, maximal ratio combining and space time coding techniques.…”

The impact of fading and interference cancelation on node density - Outage probability tradeoff in wireless networks

“…Based on this model and ignoring the effect of fading, Sousa [2] has obtained the characteristic function (CF) of the aggregate (total) interference at the receiver, which can be transformed into a closed-from probability density function (PDF) in some special cases, and, based on it, the error rates for direct sequence (DS) and frequency hopping (FH) systems. For such a model, the distribution of the distance to nearest (or k-th nearest) interferer and, thus, of its interference power can be found in a compact closed form [9]- [11], [13].…”

“…A different approach has been adopted in [9]- [11]: the power of the nearest (dominant) interferer is used to evaluate the performance instead of the total interference power. As a result, closed-form analytical performance evaluation becomes straightforward and significant insight can be extracted from such a model, including the scenarios where most powerful interferers are cancelled, either via linear or nonlinear filtering techniques.…”

“…In this paper, we rely on the approach developed in [9]- [11] and, using the methods of functions of regular variations, show that the aggregate (total) interference is dominated by the nearest interferer in the region of low outage probability (i.e. the practically-important region; see Theorem 1, 2 and 3) and, thus, both models give the same results, including the scenarios with partial/complete interference cancellation and when fading (form a broad class of distributions) is present.…”