Probabilistic flow-based spreading geographic routing for wireless sensor networks.

摘要

Technological advances are making the deployment of wireless sensor networks (WSN) a reality in many fields. Once deployed, however, WSNs usually have limited energy resources, which bound their performance in time. Furthermore, even though sensors may cover extensive regions, it is the effective report of detected phenomena at the base-station that constitutes satisfactory performance, i.e., effective spatial coverage. Given the high energy cost of wireless communication, and the expected large sizes of WSNs, one key component that affects the performance of the WSN in both the domains of time and space is the routing infrastructure. Previous routing solutions provided for wired or mobile ad-hoc networks are unsuitable for WSNs, for they do not operate under the same characteristics.;In this dissertation we propose Probabilistic Flow-based Spreading (PFS) geographic routing protocol to improve the performance of WSN applications. Our protocol probabilistically forwards packets down multiple paths based on localized flow information. In addition, it reacts to dynamic traffic conditions so that traffic will not overload at any point, lowering the probability of congestion. When congestion is detected, we adopt Probabilistic Fair Queueing (PFQ), which selectively drops packets based on flow so that all sources have their fair share of bandwidth to report events at the base-station.;We also present a theoretical analysis on how much network lifetime can be increased when packets are forwarded through multiple paths. We develop closed-form equations that allow us to compute the expected network lifetime when link-level retransmission is negligible, and the probability distribution function of the network lifetime otherwise. In order to measure the performance in both domains of time and space we propose the metric of Network Lifetime x Packets Received (TP). We show through simulations under different application scenarios that our approach outperforms greedy forwarding by 45%, maximum energy by 35%, spreading traffic based on energy or distance by 25% and GPSR by at least 20% when evaluating under the TP metric. Also we show that running PFQ on top of PFS when congested improves the per source fairness in terms of source level throughput.