Randomized Data-Gathering protocol for time-driven sensor networks

Abstract

Proactive or time-driven sensor networks are devoted to the continuous reporting of environmental data to a sink or base station. An important issue associated with these networks is the management of a potentially large number of packets that are regularly generated by the set of nodes. A common solution at the MAC level is to use scheduled TDMA-based protocols, which minimize the communication duty cycle. However, TDMA schemes have strong synchronization requirements and exhibit low adaptability to changing traffic conditions. Thus, this paper proposes an alternative MAC protocol that overcomes the limitations of TDMA-based protocols while still approaching their performance in terms of lifetime. Essentially, this proposed solution combines randomization of the sensing and transmission process with a mechanism based on transmission announcements between adjacent nodes. The result is a lightweight protocol, called Randomized Data-Gathering (RDG), which exhibits desirable characteristics with respect to energy efficiency, synchronization avoidance, performance, scalability and adaptability to traffic changes in both time and space. The role of the randomization distribution is considered in detail, and an effective randomization scheme is selected. Both analytical and simulation methods are applied.

Introduction

In proactive, time-driven or continuous monitoring wireless sensor networks (TD-WSN), communication is triggered by nodes that regularly deliver sensed data to a base station [1], [2], [3]. These data can be used for scientific purposes and/or for undertaking appropriate control tasks depending on the application that is being supported. A common scenario is the continuous reporting of some ambient condition (e.g., temperature, humidity, or light) to a data management center, which can be a simple PC connected to the base station via the Internet. Various examples are provided in [1], [4], [5], [6].

It is also common for time-driven sensor networks to be deployed in a structured manner, either by selecting strategic locations or by adopting a regular sampling pattern [3], [4], [6], [7]. Moreover, because the strategic locations can be far from one another or because the monitored variables can exhibit low spatial variability (as is usually the case), the resulting deployment tends to be sparse with a relatively low number of nodes (from tens to at most hundreds of nodes). Also, as stated in [3], [6], data in structured networks are typically routed through multi-hop, pre-determined paths that form a tree rooted at the base station or, more specifically, a reverse-multicast or convergecast structure called the data gathering tree [4], [6], [8].

On the other hand, the traffic generated by proactive networks usually takes the form of a continuous and predictable flow. Therefore, contention-free scheduled MAC protocols, such as those based on TDMA, become especially appropriate in these networks [1], [4]. Moreover, as pointed out in [1], [4], an additional advantage of TDMA schemes for sensor networks is that nodes can perfectly schedule sleep periods whenever they are not committed to transmit or receive. In this way, their duty cycle can be reduced to the minimum assigned workload. However, TDMA schemes also suffer from significant drawbacks: the need for synchronization, the need for channel (slot) assignment without conflicts, and low adaptability to changing traffic conditions and variations in the number of nodes.

The protocol proposed in this paper, called Randomized Data-Gathering (RDG), has been designed to overcome the drawbacks of TDMA schemes while still preserving their reduced average duty cycle and high lifetime expectancy. In some sense, the goal is to provide an “asynchronous” version of TDMA for time-driven sensor networks. Basically, this goal is accomplished through a two-step process. First, each sensor node is allowed to sample the environment and transmit the corresponding data at internally-generated random instants (jittered sampling process). Next, in order to obtain a reduction in the duty cycle that is similar to TDMA schemes, a second mechanism that is based on transmission announcements between adjacent nodes is introduced. The result is a lightweight MAC protocol that exhibits desirable characteristics in terms of energy efficiency, synchronization avoidance, scalability and adaptability to traffic changes in both time and space. Its applicability spans a wide variety of environmental monitoring scenarios except those requiring high sampling rates and/or small jitter levels.

The rest of the paper is organized as follows. In Section 2, related work is addressed. In Section 3, the RDG protocol is described in detail. Next, a simplified configuration of this protocol is proposed in Section 4. Based on this RDG configuration, two analytical models are constructed in Section 5. One model focuses on network lifetime, and the other model allows the determination of the average reporting frequency that satisfies certain delivery requirements. Next, in Section 6, RDG is compared to TDMA via simulation. The key measured outcome in this analysis is network lifetime. In particular, the role of the randomization distribution is analyzed in detail, and an effective randomization (distribution and parameters) is selected. The results obtained in this section also serve to validate the analytical models that are proposed in the previous section. In Section 7, other performance metrics, such as path delay, packet loss rate, throughput or jitter, are taken into consideration. In this context, it is observed that large packet error rates may cause significant lifetime degradation; thus, in these cases, an energy-saving mode (LPL) is introduced into the protocol, yielding satisfactory lifetime results. In Section 8, a scalability analysis is conducted. Finally, the main conclusions are presented in Section 9.