With the spectacular progress of technology, we have witnessed the appearance of wireless sensor networks (WSNs) in several fields. In a hospital for example, each patient will be provided with one or more wireless sensors that gather his physiological data and send them towards a base station to treat them on behalf of the clinicians. The WSNs can be integrated on a building surface to supervise the state of the structure at the time of a destroying event such as an earthquake or an explosion. In this paper, we presented a Mobility-Energy-Degree-Distance to the Base Station (MED-BS) Clustering Algorithm for the small-scale wireless Sensor Networks. A node with lower mobility, higher residual energy, higher degree and closer to the base station is more likely elected as a clusterhead. The members of each cluster communicate directly with their ClusterHeads (CHs) and each ClusterHead aggregates the received messages and transmits them directly to the base station. The principal goal of our algorithm is to reduce the energy consumption and to balance the energy load among all nodes. In order to ensure the reliability of MED-BS, we compared it with the LEACH (Low Energy Adaptive Clustering Hierarchy) clustering algorithm. Simulation results prove that MED-BS improves the energy consumption efficiency and constructs a stable structure which can support new sensors without returning to the clusters reconstruction phase.
During recent years, we have seen a miniaturization brings technology. This aptitude for the miniaturization brought a new generation of telecommunication networks which presents important challenges. The Wireless Sensor Networks are one of the technologies aiming at solving the problems of this new telecommunication and computer age [
The WSNs are composed of a large number of nodes communicating between them and distributed on a given geographic zone to measure a physical quantity or to supervise an event (temperature, pressure, earthquake···) ([2-5]).
The WSNs architecture breaks up into three underlayers: a sensor network which is composed of the information received from the external world, the clusterheads witch carried out the complex tasks of signal treatment, and a base station which received the information on behalf of the clusterheads ([5-7]).
The WSNs are particular networks, having different characteristics from the wired networks (absence of infrastructure, resource’s constraints, heterogeneity and dynamics structure). So, it was necessary to think of an auto-organized virtual topology which should be adaptive and effective in energy ([8,9]).
To conceive such topology, several solutions were suggested in the literature like the clustering, the heterogeneous networks and the dorsal.
In this paper, we proposed a clustering algorithm adopted with the small-scale Wireless Sensor Networks which goal is the minimization of the energy consumption by taking on account the patient’s mobility.
In this section, we will classify the clustering algorithms according to whether the deployed nodes are homogeneous (have the same features) or heterogeneous.
Heterogeneous network is a network where certain nodes have more raised capacities (processor, capacity for treatment, power of transmission, band-width, power of energy,···) than others. The use of heterogeneous networks can triple the delivery average rate and offer a network lifetime five times larger than the homogeneous networks [
Several algorithms of heterogeneous networks were invoked in the literature such as [
The authors introduce another algorithm of the heterogeneous nodes in [
Another algorithm was proposed in [
Others authors proposed the DEEC (Distributed Energy Efficient Clustering) [
Admittedly, this algorithm is effective and makes nodes share energy consumption between, but the enormous exchange of the control messages can involve a performance degradation of the structure.
MDC/PEQ (Mobile data collector/PEQ) [
Another example of the clustering heterogeneous nodes’ algorithm is found in [
Cheick Tidjane KONE introduced into [
Several researchers such as [
In 1995, the HCC (Highest Connectivity Cluster) [
In 2000, Amis and al. introduced the algorithm max-Min d-cluster [
Another example of clustering is found in [
Several improvements were made to LEACH such as LEACH-C [
In 2002, Chatter and al. introduced an algorithm based on the principle of LCA called WCA (weighted Clustering Algorithm) [
Mitton and al. proposed a clustering algorithm in [
HEED (Hybrid Energy-Efficient Distributed) [
DEBC [
In 2008, Yu and al. introduced the EEDMC [
The sensors networks are used for vital and crucial applications (monitoring of habitat, detection of earthquake, military monitoring, ···). For this reason, reliability represents a very important challenge ([5,6], [29,30]). In addition, energy consumption (lifetime of network) presents the most important metric in the performance evaluation of network [
Then we propose an efficient algorithm to solve this problem. Some of the designed goals are:
• Minimize the quantity of data transmitted in the network.
• Define a standby mode
• Balance the energy dissipation between nodes.
• Limit the number of hops between an ordinary node and clusterhead to 1 hop.
• When CH receives the data, it transmits them directly towards the base station.
• Reduce the nodes transmission range.
• Minimize the number of control messages in the clusters construction phase.
The sensor network considered is composed of three levels (
The whole clusterheads (CH) constitute the second level; they merge the attentive messages transmitted by their members and send the created signal towards the base station. CH and its members form a cluster. A cluster is defined as being the coverage area of its CH.
The members as well as CHs have the same features (a limited battery in energy and two radios receiving/ transmitters: to communicate with the network of the first level and another for the communication with the base station). The base station forms the third level, it treats the received data.
The members of cluster communicate directly with their CHs (connectivity intra-cluster to 1 hop). CHs communicate directly with the base station. This communication procedure is defined in LEACH clustering algorithm [
We suppose that the nodes of the first level work on the frequency channels 802.15.4 (zigbee). Indeed, four frequency channels are enough to sweep all the communication surface while being based on the principle of frequency re-uses. We also suppose that CHs use the protocol pile of standard 802.11 for their communication with the base station.
We will model the RCSF by a graph where V represents the whole of sensors and
represents the whole of wireless connections between nodes. R is the communication range, and D(U,v) defines the Euclidean distance between the nodes U and V.
The properties are the following:
• N: the number of nodes.
• ID (U): the identifier of node U.
• D (U): the connectivity degree of U.
• M (U): the mobility of U.
• Ec/com(U): power consumption by communication unit of U.
• Dis(U): the distance between the node U and the base station.
• Neigh (U): is the whole of nodes in the neighborhood of 1-hop of U.
• D (U): the degree U.
• Weight (U): the weight of U.
• State (U): state of U. We distinguish two states: “CH” (clusterhead) and “Nm” (nodes member)
• T: is the period of standby mode (deactivation period).
In this section, we propose a new clustering algorithm called MED-BS (Mobility Energy Degree Distances to Base Station) Clustering Algorithm for the sensors networks of which the goal is the minimization of the power consumption in the cluster creation phase.
Mobility is the leading cause of topology changes in the sensors networks. It should be essential to integrate mobility metric for the clusterheads election and the clusters’ formation.
We will define three mobility levels for sensors:
• Level 1: nodes speed is very weak in this case, speed lies between 0 and 5 km/h.
• Level 2: nodes speed is average, in this case, speed lies between 5 km/h and 20 km/h.
• Level 3: nodes speed is high, in this case, speed lies between 20 km/h and 44 km/h.
We suppose that the sensors speed is constant. The sensor mobility is characterized by the mobility level and can have the following values:
• M(U) = 1, if the node speed U belongs to first level.
• M(U) = 2, if the node speed U belongs to second level.
• M(U) = 3, if the node speed U belongs to third level.
We also suppose that sensors nodes know in advance their mobility levels and that the nodes having mean and high mobility will not take part in the clusterheads election phase. The purpose of this assumption is to maintain the stability of the structure. The consumed power by the mobility of nodes is not considered into account.
The energy consumption rate in the sensors networks represents the most important metric in the perform ances’ evaluation phase. This parameter depends on the used nodes’ characteristics (standby mode, nature of data processing, transmitted power, ···), and nodes behavior during the communication (retransmission, congestion, diffusion of the messages, ···) [
The consumed power by sensor is that the consumed power by these capture units, treatment units and communication units. So the energy consumption formula is defined as follows [
where:
• Ec/capture: is the energy consumed by a sensor during the capture unit activation. This energy depends primarily on the type of detected event (image, its, temperature···) and of the tasks to be realized by this unit (sampling, conversion ···).
• Ec/treatment: is the energy consumed by the sensor during the activation of its treatment unit.
• Ec/communication is the energy consumed by the sensor during the activation of its communication unit.
The consumed energy by sensors during communication is larger than those consumed by the treatment unit and the capture unit. Indeed, the transmission of a bit of information can consume as much as the execution of a few thousands instructions [
The communication energy breaks up into emission energy and reception energy:
Referring to [
where:
• K: message length (bits).
• D: distance between transmitting node and receiving node (m).
• λ: of way loss exhibitor,.
• Eelec: emission /reception energy,.
• εamp: transmission amplification coefficient, .
In [
We suppose that the aggregation energy cost respects the limiting value introduced into [
where EDA: power consumed during aggregation.
Step 1: Each node sends a message “hello” for the discovery of 1-hop neighborhood.
Step 2: Nodes having a low level of mobility (M(U) = 1) calculate their weights, the weight is calculated as follows:
Two nodes do not having the same weight because of the distance parameter (Dis(U)).
Step 3: The nodes diffuse their weights towards their neighbors.
Step 4: The node which has the weakest weight is declared like clusterhead by putting its state = “CH” and sends a message “clusterhead_elected” (containing its identity) to its neighbors.
Step 5: The neighbors receiving this message, declare themselves like “Nm”, send to the clusterhead a message “clusterhead_accepted”, and record the identity of their clusterheads in their databases.
Condition 1: A node receiving two messages “clusterhead_elected” on behalf of both clusterheads, chooses that having the weakest weight.
Condition 2: A node having a worthless degree (not having neighbors), sends its data directly towards the base station and starts the “to join a new cluster” procedure (this procedure will be thereafter detailed).
Condition 3: An outgoing node (from the cluster), sends its data directly towards the base station and starts the “to join a new cluster” procedure.
Condition 4: A clusterhead checks its reserve of energy periodically, if the remaining energy is about 40% * initial energy, then the clusterhead starts the procedure of “change clusterhead” then is declared like “Nm”.
Periodically, the base station sends to the disconnected nodes the list of clusterheads and their place. Each node calculates at each period its distances from different clusterheads, if a distance is ≤R, then it sends a message “hello” towards the concerned clusterhead. The clusterhead sends its ID and the node joins this cluster by sending a message “clusterhead_accepted”.
Step 1: The clusterhead sends to its neighbors a message “clusterhead-changes”, and is declared like “Nm”.
Step 2: The nodes having a low mobility calculate and send their weights Step 3: The node having the weakest weight is declared like “CH”, and diffuses a message “clusterhead_elected ”.
Step 4: The neighbors send to the clusterhead a message “clusterhead_accepted”, and record the identity of their clusterhead in their databases.
Each member has one period of deactivation T. It awakes each time, collects information and sends it towards its clusterhead. The clusterhead aggregates received informations and sending the built message towards the base station.
The results of our algorithm are getting using Matlab 7.0.1 in a computer Intel® Pentium® Dual CPU 1.86 Ghz with 1.99 Go of RAM.
The network of first level is composed of set of sensors. The number node in the sensor network varies between 10 and 200 nodes. The mobility of each sensor is supposed constant, and a speed is dedicated for each level of mobility: level 1:1 km/h, level 2:5 km/h and level 3:20 km/h. The initial energy for each sensor is equal to 0.5 J.
The simulation of our algorithm was carried out during 10 deactivation intervals T (standby mode) in a space of 150 m × 150 m and the range of the nodes (Tx-Arranges) varies between 20 m and 100 m. The size of a measured data package for sensors and envoy towards their clusterheads is 4000 bits.
During simulation, several metric were taken into account: the energy consumption, median number of clusterheads, median number of emitted packages towards the base station, average number of emitted packages towards the clusterheads, control traffic emitted/received during the clusters construction phase and the control traffic emitted/received during the data emission phase.
In this section, we will represent the results of our algorithm by varying the nodes range then we will compare our algorithm with LEACH algorithm while varying the size of the network each time.
In follows, we consider 100 nodes spaced in a geographical zone of 150 m × 150 m, the range of the nodes varies between 20 m and 100 m.
The following figure (
On the same figure, we can see that the percentage of the consumed power remains weak (about 0.088%) and does not exceed the 0.146% in the worst cases (100 m), these values remain reasonable for a network having 100 nodes.
The following figure (
heads) on tx-range. The number of packages sent towards the clus-terheads increases regularly with tx-ranges, that is due to the amplification of the clusterheads members by multi-plying their coverage areas.
The same figure shows that the shape of the second curve is opposed to that of the first. Indeed, the increase in the tx-range minimizes the number of the clusterheads created (
The same figure shows us after the clusters construction phase, the control traffic decreases according to the tx-range, the measured values are rather lower than those measured with the first phase. This pace shows well the stability of built structure throughout the data sending phase.
The same pace characterizes the received control traffic (
Among the most known clustering algorithm in literature: we distinguish, LEACH algorithm. LEACH is a famous
algorithm which goal is the minimization of the energy consumption in the sensors networks.
We wish in this part to compare MED-BS Clustering Algorithm with the LEACH clustering algorithm. The same energy and mobility models were considered for the two algorithms. LEACH was carried out during 10 successive towers, in parallel; MED-BS was carried out during 10 successive deactivation periods. The networks’ size tested varied between 10 and 200 nodes.
The same figure shows that MED-BS Clustering Algorithm produces less clusterheads in most shared of cases (size between 40 and 200). For LEACH more clusterheads are necessary for a larger cardinality. For MED-BS, the same number of clusterheads can be used to manage a higher network size. That explains the effecttiveness of the structure created by MED-BS if a set of sensors is added.