ABSTRACT:  While the number of wireless devices and their demanded throughput is ever increasing, the wireless spectrum remains a constant, scarce resource. The total in-band interference power is the limiting effect in highly-frequency-reused wireless systems. Therefore the study of the statistics of the total interference from very many interferers (sensor network, ad-hoc/mesh/cooperative multihop network, cellular uplink,...) becomes increasingly important. In fact, we argue that this may be the most important problem in interference analysis today, since fast simulation hardware can quickly process smaller scenarios, whereas very large scenarios often exceed time and/or machine memory. On the other hand, analysis for asymptotically large N may be easier than for finite N. Our propagation model involves power-law pathloss and lognormal shadowing, and N interferers with an arbitrary geographical distribution with respect to a receiver suffering the interference. A simple analysis will show that while assuming independent shadowing paths may not be problematic for small N, for large N the total interference statistics are very different. For this reason, it is imperative to use a realistic shadowing spatial correlation model. Since correlation models are today still quite crude, we develop techniques that allow for any (valid) correlation model. As models become more refined, our analysis will remain applicable. We show previous work where we were successful into finding the interference distribution to a large accuracy for the case when the receiver is outside the network of interferers. We showed (comparing analysis directly with simulation) that for high N the total interference is approximately lognormal, which is counterintuitive, because it is usually not the case for small N. The calculation of the lognormal parameters is as simple as possible in this case. We believe that we have the simplest possible (approximate) solution to this very involved problem. However, we found that our method worked very poorly when the receiver was within the area of interferers. We are now working on this problem, which has several more complications than the first. We show how the problem can be reformulated in various ways, each using a different mathematical branch. We also discuss how we may efficiently simulate these very large scenarios. Finally, we discuss how our research on simulation of this problem helps us understand its analysis, and vice versa.

 

SPEAKER: Sebastian Szyszkowicz completed his B.A.Sc. in Electrical Engineering Summa cum Laudeat the University of Ottawa, Canada in 2003, and his M.A.Sc. on the topic of Cellular Communications at Carleton University in 2007 under the direction of Prof. Halim Yanikomeroglu.  He is currently starting his third year of Ph.D. studies under Prof. Yanikomeroglu centred around the topics covered in today's talk. His Masters and Ph.D. work is sponsored by the prestegious national NSERC Canada PGS-M and PGS-D grants over four years. Mr. Szyszkowicz has given several talks on his Ph.D. work at various research institutes in Europe and Canada. Please visit his website for more information: www.sce.carleton.ca/~sz