A probabilistic and opportunistic flooding algorithm in wireless sensor networks

Abstract

In wireless sensor networks, many communication protocols and applications rely on flooding for various networking purposes. Prior efforts focus on how to design efficient flooding algorithms; that is, they seek to achieve full reliability while reducing the number of redundant broadcasting across the network. To achieve efficient flooding, most of the existing protocols try to reduce the number of transmissions, which is decided without considering any online transmission result. In this paper, we propose a probabilistic and opportunistic flooding algorithm that controls rebroadcasts and retransmissions opportunistically. It seeks to achieve a target reliability required by an application. For this purpose, it makes a given node select only the subset of its one-hop neighbors to rebroadcast the same message. It considers node relations such as link error rates among nodes in selecting eligible neighbors to rebroadcast. The sender controls the number of retransmissions opportunistically by tracking the current status of message reception at its neighbors. Simulation is carried out to reveal that our proposed scheme achieves the given target reliability with less overhead than other flooding algorithms in most cases, thus prolonging the network lifetime.

Introduction

Flooding is one of key mechanisms that are widely used in various wireless networks. It propagates a message throughout a network for various purposes. Especially, flooding is usually leveraged to establish a route to the destination for unicast routing (e.g. AODV [1], DSR [2]). Similarly, when a node should inform other nodes of its link state, its latest link information is flooded across the network (e.g. OLSR [3]). Due to its viability, diverse flooding algorithms have been proposed in various wireless networks including wireless sensor networks (WSNs).

Since the objective of flooding is to make it sure that all the nodes in a network receive the same message, flooding is generally performed by making all the nodes rebroadcast the received message. However, this becomes inefficient as the node density increases, which is a typical case in WSNs. Another issue is that it is hard to achieve high reliability because wireless links generally suffer from high error rates. Thus, to achieve high reliability, retransmissions are often exploited. It is crucial to decide which node to rebroadcast and how many times to retransmit the message in a flooding mechanism, since the rebroadcasting of too many nodes and/or redundant retransmissions may cause traffic implosion [4], which leads to unreliability and energy inefficiency. Prior studies have proposed several flooding schemes that seek to achieve high reliability while reducing redundant traffic by controlling the number of broadcasts.¹ However, the existing approaches have not considered the effect of a transmission (or a retransmission) of a given node on the message reception by its neighbor nodes quantitatively.

Furthermore, in wireless sensor networks, the network-wide full reliability² may not always be required according to the application requirements. For example, many sensor network applications such as temperature monitoring or intrusion detection system deploy many sensors redundantly to cover the monitoring area for high reliability [5], [6]. In this situation, a sink may want to disseminate a query with partial reliability. If the sink can achieve its own purpose only with R% of sensors, it may want to disseminate the query to only R% of sensors to reduce the number of rebroadcasts and thus energy consumption. Therefore, supporting flooding with partial reliability is another important technique to prolong lifetime of the sensor network.

Some schemes aiming to provide partial reliability in WSNs have been proposed in various contexts. For example, MMSPEED [7] seeks to deliver unicast packets with partial reliability required by applications in a decentralized and probabilistic fashion. However, MMSPEED only deals with unicast flows and does not consider the reliability of flooding. In addition to the partial reliability, GARUDA [8] considers a few other semantics of reliability. For example, sensor network applications may necessitate reliable delivery to sensors such that the entire sensing field is covered, not to all the sensors in the field. ESRT [9] redefines “reliability” somewhat differently. ESRT first assumes that sensory data packets are periodically reported from sensors to the sink. During a session, depending on the level of network congestion, the sensors can adjust the reporting rate (or “reliability”) to adapt to the network traffic load. However, the meaning of reliability in ESRT is different from our definition of partial reliability in this paper. To the best of our knowledge, how to support flooding with a target reliability has been missing in the literature.

In this paper, we seek to achieve a target reliability given by an application, ranging from full to partial reliability, while minimizing the number of broadcasts in a probabilistic and opportunistic manner. In other words, each node selects the subset of its one-hop neighbors, that will rebroadcast the same message by considering link error rates among the node itself, its one-hop neighbors and its two-hop neighbors for the target reliability. After a source node or a rebroadcasting node transmits a message once, it decides to retransmit or not by estimating the locally achieved reliability probabilistically and opportunistically. We note that OLSR tries to minimize the number of transmissions in flooding by making only selected nodes (called multi-point relays) rebroadcast the message. We extend the notion of multi-point relays (MPRs) to control flooding to achieve a target reliability required by applications. That is, depending on the target reliability and the link error rates³ of neighbors, the set of MPRs of a sender will be dynamically adjusted.

The rest of this paper is organized as follows. Prior studies are discussed in Section 2. In Section 3, we propose a novel flooding algorithm, called POFA. Simulation results are shown in Section 4. Finally, Section 5 concludes this paper.