Binary error models for Wireless Sensor Networks

Error modeling is essential in the design of wireless systems, because it produces a solid, reliable and deterministic source of errors used to verify the robustness of techniques and concepts employed to achieve maximum system throughput. This paper analyzes utilization of two prominent descriptive models to real Wireless Sensor Network (WSN) traffic exhibiting non-Poisson error burst and Poisson like heavy-tailed error gap distributions. A generalized gamma distribution model and a two state Markov Modulated Poisson Process (MMPP-2) were utilized to realistically model relevant statistics of the data trace captured in a multi-scenario case under ideal and sub-optimal transmission conditions. Moreover, an improved estimate for parameterization of the MMPP-2 process is presented. The results clearly demonstrate that real binary traffic is, contrarily to the physical layer traffic, stationary over long time intervals and for multiple transmission scenarios. Neither of the applied models is sufficient in capturing both error gap and burst processes, but an abstract mixture model based on generalized gamma distribution is suggested for future research.