^{1}

^{1}

^{*}

Clustering in wireless sensor network (WSN) is an efficient way to structure and organize the network. The cluster head (CH) forms dominant set in the network responsible for the creation of clusters, maintenance of the topology and data aggregation. A cluster head manages the resource allocation to all the nodes belonging to its cluster. In this paper, we propose a novel distributed clustering approach called Hybrid Weight-based Clustering Algorithm (HWCA). HWCA considers the neighborhood, the distance from the base station combined with the consumed energy as a hybrid metric to elect cluster head. The time required to identify the cluster head does not depend on the number of node and can be computed in a finite number of iterations. Our solution also aims to provide better performance such as maximizing the life time, reducing the number of lost frames in order to satisfy application requirements. Simulation results show that HWCA improves the network lifetime and reduces the number of lost frames compared with other similar approaches.

A Wireless Sensor Network (WSN) is an ad hoc network with a large number of nodes that are smart micro- sensor able to collect, transmit data (like heat, humidity, vibration...) and convert them into digital quantities autonomously [

The concept of dividing the geographical region to be covered into small zones has been presented implicitly as clustering in the literature [

The aim of this paper is to propose a distributed algorithm to select these cluster head nodes. The election is based on a hybrid metric taking into account three parameters: the one hop neighborhood, the amount of energy consumed and the distance from nodes to the base station.

The rest of the paper is organized as follows: Section 2 presents relative works for clustering on wireless sensor networks. In Section 3, we propose the Hybrid Weight-based Clustering Algorithm (HWCA). Simulation results are presented in Section 4 while conclusions are offered in Section 5.

Literature proposes several techniques for cluster formation and cluster head selection. All solutions aim to identify a subset of nodes within the network and bind it a leader. The first solutions are based on a single metric to elect a node as cluster head. The evolution of these algorithms has proven that the combination of several metric is more efficient for better performance.

The HCC (High-Connectivity Clustering) algorithm proposed in [

The authors of [

The use of a single metric to elect cluster head is not wise to generate stability on cluster head formation. Thus, clustering algorithms that combine multiple metrics for the cluster head election have been proposed in the literature.

The Hybrid Energy Efficient Distributed clustering protocol (HEED) [

The MWBCA algorithm (Multi-Weight Based Clustering Algorithm) [

WCA (Weighted Clustering Algorithm) [

BLAC (Battery-Level Aware Clustering) proposed in [

another metric. The remaining energy of node u is defined as followed:

the initial capacity of the node battery (initially the same for every node),

BLAC-bg for Battery-Level Aware Clustering―Battery deGree is based on node degree. This variant uses a one-hop neighborhood to build the network, so it stabilizes quickly. Any single change has a direct impact on neighbors and so on degree.

BLAC-bs for Battery Level Aware Clustering―Battery denSity uses the density ρ(u). This variant computes the clustering structure with 2-hop information but the stability is improved because a single node has less impact on its neighbors.

Battery-Level Aware Clustering―RNG deGree (BLAC-rg) and Battery-Level Aware Clustering―RNG density (BLAC-rs) variants are variations of the first and the second ones. The biggest difference is that the algorithm runs in two steps. Before computing its metric (degree or density), a relative neighborhood graph is computed in order to keep only an interesting subset of nodes. This allows memory storage saving and the use of less computing capacity for the clustering computation.

HWCA is the a distributed clustering algorithm which combines a hybrid metric composed by one hop neighborhood, consumed energy and distance from base station. Yet, nodes naturally change roles over time based on value of the hybrid metric to provide better performance. HWCA also provides efficient cluster head with no predefined size in order to match the underlying network topology and to be reliable to small topology changes. Unlike solutions from literature, HWCA builds dynamic metric efficient clusters in a distributed way. Its main goal is to provide better performance (like for instance network lifetime, delivery ratio, etc.).

In order to minimize the energy consumption of each micro-sensor and to generally increase the lifespan of the network, we propose the HWCA algorithm whose originality is based on a hybrid metric combining the neighborhood, distance from nodes to the BS and the energy consumed by each node. The HWCA algorithm also provides an energy balance through the network nodes. Indeed, the nodes then change naturally cluster head according to the value of the metric. In our algorithm a clustering organization is run over the network to carry better performance. Each node sends its data to its cluster head. Once all data are gathered, the cluster head aggregates them and sends them to a base station. In this section, we present first our new hybrid metric and secondly the cluster head election algorithm based on this metric.

We use for cluster head election of a node u a combined weight metric

We also use the following notations:

N the maximum number of nodes that a cluster head can handle ideally. This is to ensure that cluster head are not over-loaded and the efficiency of the system is maintained at the expected level.

Therefore we defined the combined metric for each node u as:

For easily identification we use:

Equation (1) becomes:

The three system parameters

A node has better chance to become cluster head if it’s metric

After values of all the components are identified, we compute the weighted factors

Based on the preceding calculation of metric

We model a wireless sensor network as a graph

The goal of Algorithm 1 is to determine for each node u its cluster head. Once the metric of each node is computed, HWCA runs algorithm 1 at each node as follow: initially, each node u is its own cluster head and there are no other nodes that have chosen u as cluster head (instruction 1 to 3). Then, for each node v in the neighborhood of u, if both metric of u and metric of cluster head of u is higher than metric of v (instruction 4 to 12) and node u is not cluster head of another node (instruction 13), node u chooses v as cluster head (instruction 14).

In our algorithm we can notice that if a node doesn’t have any other node in its neighborhood, it stays as cluster head since the initialization phase and send directly his data to the base station. When a cluster head receives data from a regular node, it stores them until it needs to send its own data and then sends the aggregated data to the base station.

We demonstrate our weighted clustering algorithm with the example shown in

We suppose that after a certain time, node y and v die. The cluster head reelection procedure is depicted in

To evaluate the performances of HWCA, we perform some simulations under the NS-2 simulator (NS-2.351). We compare HWCA to LEACH because of its energy efficiency concern. In order to observe different behaviors for LEACH, two values of the p parameter are used 5% and 10% (p is the average number of cluster head in the network).

Algorithm 1. HWCA algorithm runs at each node u.

Node | CH | Explication |
---|---|---|

v | z | Node v selects z as CH because it has the lowest metric from its neighborhood. Therefore v becomes a simple node. |

y | t | Node y selects t as CH because it has the lowest metric from its neighborhood. Therefore y becomes a simple node. |

x | u | Node x selects u as CH because it has the lowest metric from its neighborhood. Therefore x becomes a simple node. |

z | z | Node z becomes CH because it has the lowest metric from its neighborhood. |

u | u | Node u becomes CH because it has the lowest metric from its neighborhood. |

t | t | Even if node t does not have the lowest metric from its neighborhood (m(u) < m(t)), it becomes a CH because another node for instance y has already chose t as CH. |

Node | CH | Explication |
---|---|---|

x | u | Node x selects u as CH because it has the lowest metric from its neighborhood. Therefore x becomes a simple node. |

z | z | Node z has no more neighbors therefore it stays CH. |

u | u | Node u becomes CH because it has the lowest metric from its neighborhood. |

t | u | Node t changes its state from CH to a simple node because no node chooses t as cluster head and node u has the lowest metric from its neighborhood |

In order to use a realistic model for transmitting and receiving costs we consider the Texas Instruments CC2420 ZigBee2 wireless module as sensor network. These sensor nodes consume 0.77 mW when idle, 35.46 mW for receiving (Rx), and 31.32 mW for transmitting (Tx). Data traffic is also simulated and each node generates 256 kb/s of data and sends them to its cluster head. When a cluster head receives data from a child, it stores them until it needs to send its own data and then sends the aggregated data to its base station. The number of nodes is 20 either 50 while the initial energy of each one is set to 115 J. Nodes are placed on a 100 × 100 grid.

We define the network lifetime as the time until all nodes die [

In this paper, we have introduced a new clustering algorithm. HWCA combines the neighborhood, the distance from nodes to the BS and the energy consumed by each node as a hybrid metric. HWCA also balances energy consumption through the network as nodes dynamically change their role (from simple node to cluster head and vice-versa) depending on the value of the metric and the cluster head selection algorithm. The algorithm is distributed and modifications due to network dynamics are handled locally, allowing scalability. Results show that our proposition improves network lifetime with up to 40% and delivery ratio. For future work we will extend the comparison of HWCA with other algorithms and plan to run experimentations in real world.

Cheikh Sidy Mouhamed Cisse,Cheikh Sarr, (2015) A Hybrid Weight-Based Clustering Algorithm for Wireless Sensor Networks. Open Access Library Journal,02,1-10. doi: 10.4236/oalib.1101574