Stochastic Performance Trade-offs in the Design of Real-Time Wireless Sensor Networks

Future sensing applications call for a thorough evaluation of network performance trade-offs so that desired guarantees can be provided for the realization of real-time wireless sensor networks (WSNs). Recent studies provide insight into the performance metrics in terms of first-order statistics, e.g., the expected delay. However, WSNs are characterized by the stochastic nature of the wireless channel and the queuing processes, which result in non-deterministic delay, throughput, and network lifetime. For the design of WSNs with predictable performance, probabilistic analysis of these performance metrics and their intrinsic trade-offs is essential. Moreover, providing stochastic guarantees is crucial since each deployment may result in a different realization. In this paper, the trade-offs between delay, throughput, and lifetime are quantified through a stochastic network design approach. To this end, two novel probabilistic network design measures, quantile and quantile interval, are defined to capture the dependability and predictability of the performance metrics, respectively. Extensive evaluations are conducted to explore the performance trade-offs in real-time WSNs.