Overflow of fountain codes in multi-hop wireless sensor networks

Abstract

Abstract-This paper concentrates on the proper use of fountain codes for the transmission of sporadic data in a wireless sensor network (WSN). Fountain codes offer great perspectives for the self-organization of WSNs: they self adapt to the channel error rate without control packets. Deploying fountain codes in a WSN raises two problems. First, the size of the data transmitted by a sensor is small in comparison to the size usually considered with fountain codes. Second, WSNs mostly rely on multi-hop transmissions. It implies a non null transmission duration for the end-to-end acknowledgement of the reception. During this period of time, the source is still transmitting useless packets, creating a specific overhead we define as the overflow. This paper brings the overflow problem to light and analyses its impact on the network performance. Our work can be viewed as the networking counterpart of the results presented by Pakzad et al.at ISIT 2005 applied to WSNs.

On the number of transmissions vs. redundancy tradeoff for flooded fountain codes

“…In this context, Fig. 5 shows a set of curves that plot the interplay between T res avg and 1 − P fail (q res r ) for several values of N res and E, and for Note that because the priority is to minimize T res avg , increasing θ leads to an increase of E rather than N res , because the latter typically leads to a much larger number of transmissions for fixed E. In fact, we have c 0.94 = (3, 7), c 0.96 = (3,8), and c 0.98 = (3,9), and only for θ = 0.99 does c 0.99 = (4,9) involve the increase of N res . This conclusion suggests that, in the presence of a performance threshold corresponding to reasonable reliability (e.g., θ ≥ 0.9), incrementally better improvements are achieved by simply increasing the amount of redundancy (which requires only the source to act), without involving extra nodes in the flooding process (which would instead require extra coordination among the nodes).…”

“…For example, the SYNAPSE++ protocol [3] employs a fountain code to reliably disseminate firmware to a wireless sensor network. Fountain codes have also been shown to improve the performance of unicast communications, provided that early acknowledgment packets are sent in order to prevent the generation of useless redundancy [8]. In [6], the introduction of fountain codes in vehicular networks leads to a significant performance improvement compared to a simpler data carousel dissemination approach.…”

“…However, multihop communications take place through a single, fixed path known a priori, rather than considering a flooding scenario as we do in this paper. The work in [8] also focuses on a fixed path in a wireless sensor network. The authors measure the complexity of four fountain code relaying strategies, and evaluate the impact of the unrequired redundancy packets sent while the network waits for end-to-end ACKs to reach the source.…”

Energy Consumption with Fountain Codes and Clustering in Wireless Sensor Networks: Survey & Analysis‎

“…The global acknowledgement is the second most important feature of fountain code in comparison with other protocols such as Automatic Repeat reQuest (ARQ) where each data packets needs an acknowledgement (Ack) packet. In the case of error retransmission is made only for the last sent packet, thus, greatly minimizes the number of transmissions, the use of the feedback channel and subsequently the energy consumption [5]. A reliable decoding algorithm for a fountain code is one which can recover the original K input symbols from any set of m output symbols.…”

Controlled Flooding of Fountain Codes

“…For example, the SYNAPSE++ protocol [3] employs a fountain code to reliably disseminate firmware over a wireless sensor network. Fountain codes have also been shown to improve the performance of unicast communications, provided that early acknowledgment packets are sent in order to prevent the generation of useless redundancy [8]. In [6], the introduction of fountain codes in vehicular networks leads to a significant performance improvement compared to a simpler approach where data packets are periodically retransmitted.…”