부산대학교 도서관

Underwater network simulation and performance analysis require accurate packet error models. The packet error probability depends on the packet length and the temporal distribution of bit errors. We analyze error traces from decision-feedback-equalized single-carrier acoustic communication links from several shallow-water experiments and show that clustering of bit errors occurs at several timescales. We propose a two-part statistical error model consisting of a generalized Pareto fractal renewal parent process that drives Bernoulli daughter processes with generalized extreme value distributed lifetimes. We present an algorithm to simulate communication errors using this error process model and show that the simulated packet loss probability accurately matches experimental observations.