An Improved Multi-Layered Architecture and its Rotational Scheme for Large-Scale Wireless Sensor Networks

Abstract

In this paper, we propose a highly scalable network architecture, named the Progressive Multi-hop Rotational Clus- tered (PMRC) structure, suitable for the construction of large- scale wireless sensor networks. In the PMRC structure, sensor nodes are partitioned into layers according to their distances (cal- culated using hop counts) to the sink node. A cluster is composed of the nodes located in the same layer and within the transmission range of the cluster head which is located in one layer up. Each cluster here actually selects two cluster heads, which makes the PMRC structure different from another multi-layered structure, MINA (4). Based on the observation that load balancing tends to help balance the energy consumption among different sensor nodes and consequently prolong the network life time, we further propose a rotational scheme functioning at two levels: 1) the two cluster heads in the same cluster rotate to receive and forward data, and 2) clusters at the same layer rotate to sense data. Exten- sive simulations have been conducted to verify the rotation scheme with two selection strategies each specially tailored for one of the two cluster heads required in the PMRC structure. These results have confirmed that the PMRC structure and its rotation scheme together can significantly prolong the node life time and reduce the number of network reconstructions compared with those obtained from a multi-layered structure with single cluster head. I. INTRODUCTION The benefits of low-cost, rapid deployment, self-organization capa- bility and cooperative data-processing have made the wireless sensor networks a practical solution for a wide range of application areas, including military, industry and commercial, environment, health and home (2), (3). The most significant challenge in sensor networks is to overcome the energy constraint since each sensor node has limited power ( ) and it is hard to replenish the power es- pecially in hazardous or hostile application scenarios. The other chal- lenge faced by sensor networks is scalability. Many applications, such as military surveillance and habitat monitoring, require the deploy- ment of large-scale sensor networks (with the number of sensor nodes in the order of hundreds or thousands, or even millions) in a large ge- ographic area, and seamless connectivity to existing infrastructures is usually required when new nodes are added. Other research work for large-scale sensor networks include (7) and (10). In (7), the SAFE protocol was proposed for data dissemination from stationary sensor nodes to mobile sink nodes in large-scale sen- sor networks. The major problems of the SAFE protocol are the large number of states to be maintained at intermediate nodes and the mul- tiple rounds of message exchanges required to set up a path. The two- tier data dissemination (TTDD) protocol (10) is another protocol for disseminating data from stationary sensor nodes to multiple mobile sinks by setting up a grid structure. However, the cost of proactively creating/maintaining the grid structure from all sources to the edge of the sensor field tends to be unbearably high for large sensor networks. In this paper, we follow the layered structure and subsequently pro- pose the Progressive Multi-hop Rotational Clustered (PMRC) struc- ture as an extension to the MINA structure (4). In a PMRC structure, a cluster is formed in the way similar to that in MINA but with a signif- icant distinction: here two cluster heads are selected for each cluster. To balance the load and the energy consumption among different sen- sor nodes, we propose a rotational scheme functioning at two levels: 1) the two cluster heads in the same cluster rotate to receive and forward data, and 2) clusters at the same layer rotate to sense data. Through simulations, we show that the PMRC structure together with its ro- tational scheme outperforms the multi-layered structure with single cluster head in node life time and hence reduce the number of network reconstructions. The rest of the paper is organized as follows. In Section II, we will describe the PMRC structure. In Section III, the problems and algo- rithms of selecting the primary cluster head and the secondary cluster head are discussed. In Section IV, simulation results are presented and discussed. Section V concludes the paper.