Research
Areas RESEARCH
|
• simulation
of communication systems and networks |
|
• Summary : In digital communication systems the probability of a bit error is usually on the order of 10E-6 or so. To simulate an event in order to find its probability, you need to "hit" that event several times in order to get a good estimate of it. Typically then, for rare events, one might need to generate 10E8 to 10E9 independent random variables. This imposes severe if not impossible constraints on the random number generators commonly available. We are studying methodologies whereby the probabilistic systems under consideration are "twisted" into new ones where the events under study are much more likely to occur. One then " untwists " the estimates obtained here to derive the desired estimate for the original system. Orders of magnitude better performance can be obtained from these twisted simulations. |
| *
Project A
We are currently employing large deviation theory techniques to the study of the simulation of error events in digital receivers with feedback structures. We have found that by using proper biasing techniques we are able for the first time to simulate the error events in a decision feedback equalizer in an efficient manner. This receiver is particularly interesting since the feedback elements are nonlinear thus making a closed form solution to the error rate virtually impossible . |
| *
Project B
In this project we are studying system collapse in distributed power systems due to line failures. It involves a novel type of network simulation using ideas from Markov Chain Monte Carlo. Using these techniques we are trying to quantify in a topological sense the security of a power network |
Projects