**3. Heterogeneous wireless sensor networks**

This section presents a paradigm of heterogeneous wireless sensor network and discusses the impact of heterogeneous resources (Yarvis ,2005)(V. Katiyar, 2011)

#### **3.1. Types of heterogeneous resources**

There are three common types of resource heterogeneity in sensor nodes: computational heterogeneity, link heterogeneity and energy heterogeneity.

**Computational heterogeneity** means that the heterogeneous node has a more powerful microprocessor and more memory than the normal node. With the powerful computational resources, the heterogeneous nodes can provide complex data processing and longer-term storage.

**Link heterogeneity** means that the heterogeneous node has high bandwidth and longdistance network transceiver than the normal node. Link heterogeneity can provide a more reliable data transmission.

**Energy heterogeneity** means that the heterogeneous node is line powered or its battery is replaceable.

Among above three types of resource heterogeneity, the most important resource heterogeneity is the energy heterogeneity because both computational heterogeneity and link heterogeneity will consume more energy resource.

If there is no energy heterogeneity, computational heterogeneity and link heterogeneity will bring negative impact to the whole sensor network, i.e ., decreasing the network lifetime.

### **3.2. Impact of heterogeneity on wireless sensor networks**

Placing few heterogeneous nodes in the sensor network can bring following benefits:

**Decreasing latency of data transportation**: Computational heterogeneity can decrease the processing latency in immediate nodes and link heterogeneity can decrease the waiting time in the transmitting queue. Fewer hops between sensor nodes and sink node also mean fewer forwarding latency.

**Prolonging network lifetime**: The average energy consumption for forwarding a packet from the normal nodes to the sink in heterogeneous sensor networks will be much less than the energy consumed in homogeneous sensor networks.

**Improving reliability of data transmission**: It is well known that sensor network links tend to have low reliability and each hop significantly lowers the end-to-end delivery rate.

With heterogeneous nodes, there will be fewer hops between normal sensor nodes and the sink. So the heterogeneous sensor network can get much higher end-to-end delivery rate than the homogeneous sensor network.

#### **3.3. Performance measures**

318 Energy Efficiency – The Innovative Ways for Smart Energy, the Future Towards Modern Utilities

demonstrate that CHR performs better than directed diffusion and SWR.

assumed that all the sensor nodes are uniformly distributed.

LEACH and EECHE in terms of network lifetime.

**3.1. Types of heterogeneous resources** 

**3. Heterogeneous wireless sensor networks** 

heterogeneity, link heterogeneity and energy heterogeneity.

the impact of heterogeneous resources (Yarvis ,2005)(V. Katiyar, 2011)

**sensor networks** 

individual sensors.

The CHR protocol partitions the heterogeneous network into clusters, each being composed of L-sensors and led by an H-sensor. Within a cluster, the L-sensors are in charge of sensing the underlying environment and forwarding data packets originated by other L-sensors toward their cluster head in a multi-hop transmission. The H-sensors, on the other hand, are responsible for data fusion within their own clusters and forwarding aggregated data packets originated from other cluster heads toward the sink in a multi-hop transmission using only cluster heads. While L-sensors use short-range data transmission to their neighboring H -sensors within the same cluster, H-sensors perform long-range data communication to other neighboring H-sensors and the sink. Simulation results

**2.7. Energy efficient cluster head election protocol for heterogeneous wireless** 

(LI Han, 2010) proposed an energy efficient cluster head election protocol for heterogeneous wireless sensor networks and using the improved Prim's algorithm to construct an inter cluster routing. He has considered three types of sensor nodes. Some fraction of the sensor nodes are equipped with the additional energy resources than the other nodes. He has

In this protocol, the cluster head node sets up a TDMA schedule and transmits this schedule to the nodes in the cluster. This ensures that there are no collisions among data messages and also allows the radio components of each non-cluster head node to be turned off at all times except during their transmit time, thus minimizing the energy dissipated by the

In order to reduce the energy consumption of the cluster heads which are far away from the base station and balance the energy consumption of the cluster heads which are close to the base station, a multi-hop routing algorithm of cluster head has been presented, which introduces into the restriction factor of remainder energy when selects the interim nodes between cluster heads and base station, and also the minimum spanning tree algorithm has been included. The protocol can not only reduce the consumption of transmit energy of cluster head, but also the consumption of communication energy between non-cluster head and cluster head nodes. Simulation results show that this protocol performs better than

This section presents a paradigm of heterogeneous wireless sensor network and discusses

There are three common types of resource heterogeneity in sensor nodes: computational

Some performance measures that are used to evaluate the performance of clustering protocols are listed below(R. Sheikhpour et al., 2011).

**Network lifetime (stability period):** It is the time interval from the start of operation (of the sensor network) until the death of the first alive node.

**Number of cluster heads per round**: Instantaneous measure reflects the number of nodes which would send directly to the base station, information aggregated from their cluster members.


$$E\_{T\chi}(k) = E\_{R\chi}(k) = E\_{elec} \ast k \tag{2}$$

$$E\_{Tx}(k,d) = E\_{elec} \ast k + E\_{amp}(k,d) \tag{3}$$

$$E\_{T\chi}\left(k,d\right) = \begin{cases} \left(E\_{elec} \ast k\right) + \left\{\varepsilon\_{fs} \ast k \ast d^2\right\} & \text{if } d \le d\_0\\ \left(E\_{elec} \ast k\right) + \left\{\varepsilon\_{mp} \ast k \ast d^4\right\} & \text{if } d > d\_0 \end{cases} \tag{4}$$

$$d\_0 = \sqrt{\frac{\varepsilon\_{fs}}{\varepsilon\_{mp}}}\tag{5}$$

$$E\_{\mathbb{R}\mathfrak{x}}(k) = E\_{elec} \ast k \tag{6}$$

$$E\_{CH} = \left[k \ast E\_{elec} \ast \frac{N}{C}\right] + \left[k \ast E\_{DA} \ast \frac{N}{C}\right] + \left[k \ast E\_{elec} + k \ast \varepsilon\_{mp} \ast d\_{toBS}^4\right] \tag{7}$$

$$E\_{non-CH} = k \ast E\_{elec} + k \ast \varepsilon\_{fs} \ast d\_{toCH}^2 \tag{8}$$

$$d\_{\text{toCH}}^2 = \iint \left( (\mathbf{x}^2 + \mathbf{y}^2) \* \rho(\mathbf{x}, \mathbf{y}) \right) d\mathbf{x} \, d\mathbf{y} = \iint r^2 \rho(\mathbf{r}, \theta) r \, d\mathbf{r} \, d\theta \tag{9}$$

$$d\_{toCH}^2 = \rho \int\_{\theta=0}^{2\pi} \int\_{r=0}^{\left(M/\sqrt{\pi C}\right)} r^3 \, dr \, d\theta = \frac{\rho}{2\pi} \, \frac{M^4}{C^2} \tag{10}$$

$$d\_{toCH}^2 = \frac{1}{2\pi} \frac{M^2}{\mathcal{C}} \tag{11}$$

$$E\_{non-CH} = k \ast E\_{elec} + k \ast \varepsilon\_{f\_S} \ast \frac{1}{2\pi} \frac{M^2}{C} \tag{12}$$

$$E\_{cluster} = E\_{CH} + \left(\frac{N}{\mathcal{C}} - 1\right) \ast E\_{non-CH} \approx E\_{CH} + \frac{N}{\mathcal{C}} \ast E\_{non-CH} \tag{13}$$

$$E\_{\text{total}} = \mathcal{C} \, \*E\_{\text{cluster}} \,\tag{14}$$

$$\frac{\partial E\_{total}}{\partial \mathcal{C}} = 0 \tag{15}$$

$$\mathcal{C}\_{opt} = \sqrt{\frac{N}{2\pi}} \sqrt{\frac{\varepsilon\_{fs}}{\varepsilon\_{mp}}} \frac{M}{d\_{toBS}^2} \tag{16}$$

$$P\_{opt} = \frac{\mathcal{C}\_{opt}}{N} \tag{17}$$

$$\begin{aligned} N &= N^\* \, m \Big[ \text{NCG} \, nodes \, \overline{\text{J}} + N^\* \, \begin{pmatrix} 1 - m \end{pmatrix} \Big] \text{normal} \, nodes \, \overline{\text{J}}\\ E \Big[ \text{total} \, \overline{\text{J}} \right] &= N^\* \, m^\* \, E\_0 \, ^\* \left( 1 + a \right) + N^\* \left( 1 - m \right) \, ^\*E\_0 \end{aligned} \tag{18}$$

Hierarchical Adaptive Balanced Routing Protocol

(19)

(20)

for Energy Efficiency in Heterogeneous Wireless Sensor Networks 325

if S ∈ G���

� − �� (21)

requires the minimum communication energy. The duration of the steady phase is longer

Each sensor � elects it self to be a gateway at the beginning of each round, round. This decision is based on the suggested percentage of gateways for the network (determined a priori) and the number of rounds the node has been a gateway so far. The desired percentage of gateways is chosen such that the expected number of gateway nodes for each round is ��. Thus, if there are �∗� ��� nodes (advanced nodes) in the network, the

> �� <sup>=</sup> �� �∗�

Decision to become gateway is made by the node � choosing a random number � between 0 and 1. The node becomes a gateway for the current round if the number � is less than the

> P� ) ∗

We define as �(�gat) the threshold for gateway node �, � is the currenet round, �gat is set of nodes which have not been gateways in 1��g rounds , E�\_������� is the current energy of the

The main idea is for the sensor nodes to elect themselves with respect to their energy levels autonomously. The goal is to minimize communication cost and maximizing network resources in other to ensure concise information is sent to the sink. Each node transmits data to the closest cluster head and the cluster heads performs data aggregation. Assume an optimal number of clusters ���� in each round. It is expected that as a cluster head, more energy will be expended than being a cluster member. Each node can become cluster head with a probability ���� and every node must become cluster head once every 1� ����

The optimal probability of a node to become a cluster head, ����, can be computed as

�� is a number of gateway nodes, ���� is the optimum number of clusters that is expressed

���� <sup>=</sup> ����

E�\_������� E�\_������� 

0 otherwise

than the duration of the setup phase in order to minimize overhead.

��S����=� P�

node and E�\_������� is the initial energy of the node.

**5.3. Cluster head selection algorithm** 

1−P� ∗ (r mod <sup>1</sup>

**5.2. Gateway selection algorithm** 

desired percentage of gateways is:

following threshold:

rounds.

follows:

by:

**Figure 6.** The HABRP Network model

The operation of HABRP is divided into rounds. Each round begins with a set-up phase followed by a steady-state phase, as shown in Fig.7.

**Figure 7.** Time line showing HABRP operation

During the set-up phase the gateways are elected and the clusters are organized. It is constituted by gateway selection algorithm, cluster selection algorithm and cluster formation algorithm.

**Figure 8.** Time line showing set-up phase

After the set-up phase is the steady-state phase when data are transmitted from the nodes to the cluster head to agregate data and transmit it to the base station through the gateway that requires the minimum communication energy. The duration of the steady phase is longer than the duration of the setup phase in order to minimize overhead.

#### **5.2. Gateway selection algorithm**

324 Energy Efficiency – The Innovative Ways for Smart Energy, the Future Towards Modern Utilities

The operation of HABRP is divided into rounds. Each round begins with a set-up phase

Set-up Steady-state Frame

During the set-up phase the gateways are elected and the clusters are organized. It is constituted by gateway selection algorithm, cluster selection algorithm and cluster

Round 0 Round 1 …

Set-up

After the set-up phase is the steady-state phase when data are transmitted from the nodes to the cluster head to agregate data and transmit it to the base station through the gateway that

**algorithm**

**Cluster Formation algorithm** 

**Figure 6.** The HABRP Network model

followed by a steady-state phase, as shown in Fig.7.

**Figure 7.** Time line showing HABRP operation

**Figure 8.** Time line showing set-up phase

**Gateway selection algorithm Cluster Head selection** 

formation algorithm.

Each sensor � elects it self to be a gateway at the beginning of each round, round. This decision is based on the suggested percentage of gateways for the network (determined a priori) and the number of rounds the node has been a gateway so far. The desired percentage of gateways is chosen such that the expected number of gateway nodes for each round is ��. Thus, if there are �∗� ��� nodes (advanced nodes) in the network, the desired percentage of gateways is:

$$P\_g = \frac{N\_g}{N \ast m} \tag{19}$$

Decision to become gateway is made by the node � choosing a random number � between 0 and 1. The node becomes a gateway for the current round if the number � is less than the following threshold:

$$\text{T\{S\_{gat}\}} = \begin{cases} \frac{\text{P\_g}}{\text{1} - \text{P\_g} \ast \text{(r \, mod \, } \frac{\text{I}}{\text{P}\_{\text{g}}})} \ast \frac{\text{E\_{s\\_current}}}{\text{E\_{s\\_initial}}} & \text{if } \text{S} \in \text{G}\_{\text{gat}} \\\ 0 & \text{otherwise} \end{cases} \tag{20}$$

We define as �(�gat) the threshold for gateway node �, � is the currenet round, �gat is set of nodes which have not been gateways in 1��g rounds , E�\_������� is the current energy of the node and E�\_������� is the initial energy of the node.

#### **5.3. Cluster head selection algorithm**

The main idea is for the sensor nodes to elect themselves with respect to their energy levels autonomously. The goal is to minimize communication cost and maximizing network resources in other to ensure concise information is sent to the sink. Each node transmits data to the closest cluster head and the cluster heads performs data aggregation. Assume an optimal number of clusters ���� in each round. It is expected that as a cluster head, more energy will be expended than being a cluster member. Each node can become cluster head with a probability ���� and every node must become cluster head once every 1� ���� rounds.

The optimal probability of a node to become a cluster head, ����, can be computed as follows:

$$P\_{opt} = \frac{\mathcal{C}\_{opt}}{N - Ng} \tag{21}$$

�� is a number of gateway nodes, ���� is the optimum number of clusters that is expressed by:

$$\mathcal{L}\_{opt} = \sqrt{\frac{N - Ng}{2\pi}} \quad \sqrt{\frac{\varepsilon\_{fs}}{\varepsilon\_{mp}}} \frac{M}{d\_{toBS}^2} \tag{22}$$

$$P\_{nrm} = \frac{P\_{opt}}{1 + a \ast m} \tag{23}$$

$$P\_{adv} = \frac{P\_{opt}}{1 + a \ast m} \ast (1 + a) \tag{24}$$

$$\text{T(S}\_{\text{nrm}}) = \begin{cases} \frac{\text{P\_{\text{nrm}}}}{\text{1} - \text{P\_{\text{nrm}}} \ast \left(\text{r mod } \frac{\text{I}}{\text{P\_{\text{nrm}}}}\right)} \ast \frac{\text{E\_{\text{s}\_{\text{c,current}}}}}{\text{E\_{\text{s}\_{\text{initial}}}}} & \text{if } \text{S}\_{\text{nrm}} \in \text{G}\_{\text{nrm}} \\\ 0 & \text{otherwise} \end{cases} \tag{25}$$

$$\Gamma(\text{S}\_{\text{adv}}) = \begin{cases} \frac{\text{P}\_{\text{adv}}}{1 - \text{P}\_{\text{adv}} \* \left(\text{r} \bmod \frac{\text{I}}{\text{P}\_{\text{adv}}}\right)} \* \frac{\text{E}\_{\text{s}\_{\text{current}}}}{\text{E}\_{\text{s}\_{\text{Initial}}}} & \text{if } \text{S}\_{\text{adv}} \in \text{G}\_{\text{adv}}\\ 0 & \text{otherwise} \end{cases} \tag{26}$$

$$E\_{CH\_\*\text{to}\,\_{BS}} > E\_{CH\_\*\text{to}\,\_{Gat}} + E\_{Gat\_\*\,\text{to}\,\_{B\,\text{S}}} \tag{27}$$

$$E\_{CH\_{\text{-}to\\_BS}} \le E\_{CH\_{\text{-}to\\_Gat}} + E\_{\text{Gat\\_to\\_BS}} \tag{28}$$
