Discussion on LEACH performance

As you can see, LEACH tremendously improves the lifetime of the network while keeping the network distributed. Undoubtedly it is a very good protocol, because it maintains the distributiveness of the network (So that the central base station does not need to know the locations of each node).

24 | P a g e 3.2. LEACH-C (LEACH-Centralized)

While there are advantageous to using LEACH’s distributed cluster formation algorithm, where each node makes autonomous decisions that result in all the nodes being placed into clusters. this protocol renders no guarantee about the placement and or number of cluster head nodes. Since the clusters are adaptive obtaining a poor clustering set up during a given round will not greatly affect overall performance of LEACH. However using a central control algorithm to form the clusters may produce better clusters by dispersing the cluster head nodes throughout the network.

This is the basis for LEACH-C LEACH Centralized a protocol that uses a centralized clustering algorithm and the same steady state protocol as LEACH.

In addition to determining good clusters the base station needs to ensure that the energy load is evenly distributed among all the nodes To do this the base station computes the average node energy and whichever nodes have energy below this average cannot be Cluster-heads for the current round. Using the remaining nodes as possible Cluster-heads the base station runs a simulated annealing algorithm [3] to determine the best k nodes to be Cluster-heads for the next round and the associated clusters This algorithm minimizes the amount of energy the non-Cluster-head nodes will have to use to transmit their data to the Cluster non-Cluster-head by minimizing the total sum of squared distances between all the non-Cluster-head nodes and the closest Cluster-head. At each iteration the next state which consists of a set of nodes in C’ is determined from the current state the set of nodes in C by randomly and independently perturbing the x and y coordinates of the nodes c in C to get new coordinates x’ and y’. The nodes that have location closest to (x’,y’) become the new set of Cluster-head nodes c that makes up set C. Given the current state at iteration k represented by the set of Cluster-head nodes C with cost f(C) the new state represented by the set of Cluster-head nodes C’ with cost f(C) will become the current state with probability,

25 | P a g e Equation 2

where αk is the control parameter equivalent to the temperature parameter in the thermodynamic model and f() represents the cost function defined by,

Equation 3

where d(i,c) is the distance between node i and node c. The parameter αk must be chosen to be increasing with increasing αk to ensure that the algorithm converges. However, if αk increases too quickly, the system will get stuck in local minima. On the other hand, if αk increases too slowly, the system will take a very long time to converge.

3.3. TEEN (Threshold sensitive Energy Efficient sensor Network protocol) [4]

3.3.1. Functioning

In this scheme, at every cluster change time, in addition to the attributes, the cluster-head broadcasts to its members, Hard Threshold (HT): This is a threshold value for the sensed

26 | P a g e attribute. It is the absolute value of the attribute beyond which, the node sensing this value must switch on its transmitter and report to its cluster head.

Soft Threshold (ST): This is a small change in the value of the sensed attribute which triggers the node to switch on its transmitter and transmit. The nodes sense their environment continuously.

The first time a parameter from the attribute set reaches its hard threshold value, the node switches on its transmitter and sends the sensed data. The sensed value is stored in an internal variable in the node, called the sensed value (SV). The nodes will next transmit data in the current cluster period, only when both the following conditions are true:

1. The current value of the sensed attribute is greater than the hard threshold.

2. The current value of the sensed attribute differs from SV by an amount equal to or greater than the soft threshold.

Whenever a node transmits data, SV is set equal to the current value of the sensed attribute. Thus, the hard threshold tries to reduce the number of transmissions by allowing the nodes to transmit only when the sensed attribute is in the range of interest. The soft threshold further reduces the number of transmissions by eliminating all the transmissions which might have otherwise occurred when there is little or no change in the sensed attribute once the hard threshold.

3.3.2. Important Features

The main features of this scheme are as follows:

1. Time critical data reaches the user almost instantaneously. So, this scheme is eminently suited for time critical data sensing applications.

2. Message transmission consumes much more energy than data sensing. So, even though the nodes sense continuously, the energy consumption in this scheme can potentially be much less than in the proactive network, because data transmission is done less frequently.

3. The soft threshold can be varied, depending on the criticality of the sensed attribute and the target application.

27 | P a g e 4. A smaller value of the soft threshold gives a more accurate picture of the network, at the

expense of increased energy consumption. Thus, the user can control the trade-off between energy efficiency and accuracy.

5. At every cluster change time, the attributes are broadcast afresh and so, the user can change them as required.

The main drawback of this scheme is that, if the thresholds are not reached, the nodes will never communicate; the user will not get any data from the network at all and will not come to know even if all the nodes die. Thus, this scheme is not well suited for applications where the user needs to get data on a regular basis. Another possible problem with this scheme is that a practical implementation would have to ensure that there are no collisions in the cluster. TDMA scheduling of the nodes can be used to avoid this problem. This will however introduce a delay in the reporting of the time-critical data. CDMA is another possible solution to this problem.

This protocol is best suited for time critical applications such as intrusion detection, explosion detection etc.

3.4.APTEEN (Adaptive Periodic Threshold-sensitive Energy Efficient Sensor Network Protocol) [5]

In APTEEN once the CHs are decided, in each cluster period, the cluster head first broadcasts the following parameters:

Attributes(A): This is a set of physical parameters which the user is interested in obtaining data about.

Thresholds: This parameter consists of a hard threshold (HT ) and a soft threshold (ST ). HT is a particular value of an attribute beyond which a node can be triggered to transmit data. ST is a small change in the value of an attribute which can trigger a node to transmit data again.

28 | P a g e Schedule: This is a TDMA schedule similar to the one used in [APTEEN PAPER 8], assigning a slot to each node.

Count Time(TC): It is the maximum time period between two successive reports sent by a node.

It can be a multiple of the TDMA schedule length and it accounts for the proactive component. In a sensor network, close-by nodes fall in the same cluster, sense similar data and try to send their data simultaneously, causing possible collisions.

In the following section, data values exceeding the threshold value are referred as critical data.

The main features of our scheme are :

1. By sending periodic data, it gives the user a complete picture of the network. It also responds immediately. to drastic changes, thus making it responsive to time critical situations. Thus, It combines both proactive and reactive policies.

2. It offers a flexibility of allowing the user to set the time interval (TC) and the threshold values for the attributes.

3. Energy consumption can be controlled by the count time and the threshold values.

4. The hybrid network can emulate a proactive network or a reactive network, by suitably setting the count time and the threshold values.

The main drawback of this scheme is the additional complexity required to implement the threshold functions and the count time. However, this is a reasonable trade-off and provides additional flexibility and versatility.

Implementation

In order to do a fair comparison we implemented all the algorithms above in MATLAB and then used randomly generated 100 node network with some properties given.Not only we used the same network, we also employed the same network energy model as described in the following.

29 | P a g e 3.5. The simulation network

We assumed a network with 100 nodes.Nodes are deployed randomly in a 50mx50m area.The deployment is done with a random selection of x and y coordinates of each node’s location with a uniform probability.All the nodes in the network are identical in all aspects.They all are tasked with the same , have same properties related to energy. All are equipped with a battery power of 0.5J (otherwise specified), where this battery has a linear battery discharge charactersitics.Through out the network and the through out the whole duration, the environmental conditions related to energy dissipation (such has humidity and temperature) remain same.Nodes are static and do not move from the initial location (due to wind, water etc.) through out the time of the experiment.Nodes have some computational ability to run a distributed algorithms like LEACH.

Figure 8: Randomly deployed network in a 50mX50m area.

3.5.1. First order Radio Model

30 | P a g e There are different research on the energy efficient radios on the table.Also there are different assumption on how energy is spent while transmitting and receiving a chunk of data by the radio transmitter and the reciever.These depends not only the channel conditions, the radio characteristics and the Tx and Rx modes. But here we stick to a simple model which is given in [2],where the radio dissipates E_elec = 50 nJ/bit to run the transmitter or receiver circuitry and Є_amp

= 100 pJ/bit/m² for the transmit amplifier with an acceptable SNR.

No

Figure 9: First order radio model

Thus, to transmit a k-bit message a distance d using our radio model, the radio expends,

Equation 4

31 | P a g e And to receive this amount of k-bit message, it expends.

Equation 5

We make the assumption that the radio channel is symmetric such that the energy required to transmit a message from node A to node B is the same as the energy required to transmit a message from node B to node A for a given SNR. For our experiments, we also assume that all sensors are sensing the environment at a fixed rate and thus always have data to send to the end-user.

3.6.Results

Figure 10: Direct transmission (red), LEACH-C (blue)and LEACH(green) comparison no of

32 | P a g e nodes alive vs. rounds

Figure 11: no of alive nodes vs. rounds TEEN

33 | P a g e Figure 12: No of nodes alive vs. Rounds APTEEN

34 | P a g e 4. Chapter 4: Performance Metrics of WSN communication protocols

In this chapter we introduce the energy goals of the sensor networks and the parameters to measure them.

As we discussed in the first chapter, the sensor networks are energy constrained networks.

Therefore the success of a communication protocol is measured by the energy efficiency it can render to the network. So it is important to define the energy goals of the network we need to achieve and realize the metrics that measure them clearly.

First of all it is clear that the network lasting longer is desirable and must be attained. As we can understand the network capacity could be measured by the no. of alive nodes in the network. So we consider the lifetime of the network in its full capacity as an energy goal. This could mean we can take as the time of the first node dies as a metric .The similar manner we think network lifetime at any capacity is also an energy goal. Similar fashion time of the last node dies would be a performance metric. Therefore a graph of No of alive nodes Vs. Time (rounds) could give accurate picture of the protocols’ energy performance.

Since the wireless networks perform the functions collectively, it is important that all the nodes to be functioning well. When the network performs in its full capacity, service quality will be maximum. When the number of live nodes decreases the service it renders also decreases. When the service quality decreases to a some amount, the network becomes disposable even when there are nodes alive in the network.

So first node dying is more important than the last node dying. And also it is desirable that the whole network dies at the same time. That way the network will be in its fully functioning capacity for the most of the time.

Based on these facts what we need is a network works at its full performance longest time possible and then dies together so that it could be replaced by another.

35 | P a g e In order to determine if the network dies fast or not, we need to devise a metric. We take the slope of the No of alive nodes Vs. Time graph for this as it essentially reflects the network’s performance degradation rate. So what needed to be achieved is a high steepness in the graph.

Proposed performance parameter,

So after first node dies,

Equation 6

Death rate of the nodes = d (Number of nodes alive)/dt

= constant* d(number of nodes alive)/d(rounds)

= number of nodes / (Round in which last node dies - Round in which first node dies)

We also observe that in most of the improved versions of LEACH [6] [7] (though they are centralized) they use the energy balancing which would eventually achieve the same as we discussed before.(See Figure 10) Even though they do not explicitly discuss about this.

4.1. The advantages of the high steep summarized

1. Since there is a high steep all the nodes die together. So the network will be fully functioning till the first node dies, and soon after that ‘the whole network’ can be replaced instead of replacing nodes.

2. if the slope is less steep that means, at one point there will be coverage holes in the network. So the network function is not good. The only way you can fix it would be to deploy “more nodes”

at the initial deployment.

36 | P a g e But if we know the characteristics would have a steep slope that means the coverage holes will not be there for more time. (only for few rounds). So we can deploy the exact amount of nodes needed for the coverage. This is economical and makes the task easier for the network designer.

But besides, we also have another energy goal. As described in the [2],It is desirable that the network die geographically evenly. For an example five nodes dying in a close proximity to each other can reduce the network performance more than an if those five nodes were spread in a large area evenly.

37 | P a g e 5. Chapter 5 :Keeping the nodes alive

As we implied we propose a new protocol to address the energy goals we discussed above. Main goal is our algorithm is to keep all the nodes alive as much times as possible. It perform most of the communication like LEACH.LEACH is taken here not only because of its energy awareness also it is the most efficient distributed algorithm we think exists as a data gathering hierarchical protocol.

The reason why the first node dies very fast in a LEACH network is because the nodes are not aware of the energy levels of its own, and of course relative energy level w.r.t the network. Most centralized algorithms solve this problem by base station requesting the energy levels of all the nodes and then processing this data to find better clusters Cluster heads in order to optimize the energy balance of the network.

Since our objective is to retain the distributiveness of the network, we do not employ such a centralized clustering algorithm. But the nodes can send its own battery (energy) level to the base station as it sends the other information periodically. If the node represent its energy level with a 6 bit number (64 levels), the energy cost of this transmission could be considered nominal, as it is sent every 10k bit data transmission. The node also finds out its closest node (the neighbor) and it’s ID. Note that it does not need to know the location (even the relative coordinates or distance ) of the neighbor. Also during the setup phase the node finds out the energy level of the neighbor.

Also the base station sends a broadcast of the ‘average energy level of the network’. Since all the nodes send its energy level at the setup phase, BS can easily calculate this and send to all the nodes with a nominal energy cost reception.

Equipped with the information such as its own energy level, it is neighbor and its energy, the nodes go to the advertisement phase of the protocol in which it elect itself as a cluster head as in LEACH with a probability. But in our case the nodes with energy level below the average energy level, choose not to become a CH (So the P=0 for those nodes). This eliminates the risk of a low energy node becoming a CH (Which is the most probable reason of the first node to die fast.)Later the clustering of the network done in a way similar to that of the LEACH.

38 | P a g e 5.1.Data transmission phase

By this time the nodes are assigned of their work already, either to act as a CH or a normal node.

But if the node is a low energy node (i.e. energy level is below the average energy level) and the neighboring node is a high energy node, instead of data being transmitted directly to the CH it will be routed to the CH through the neighbor. If the node is a high energy node, data is transmitted directly to the CH.(Note that a low energy node cannot become a CH)

5.2. In pseudo codes Setup phase

1. Each node finds out what is its neighbor (nearest node).

2. It finds out its own energy level. (This could be a discrete level represented with 6bits) 3. send the energy level to the BS.

4. it finds out neighbors energy too.

5.Receives the average energy level of the network from the BS. (This is calculated in the base station using the information sent by all the nodes.)

Advertisement phase

5.Node determines its relative energy level.

If (the nodes energy > average energy) then Node is a ‘High energy node’

else Node is a “Low energy node”.

6.If the node is a low energy node it becomes ineligible to become a CH.

if it has high energy level, it becomes eligible. (but still randomly elects itself as in original LEACH)

7. if the node is a non CH node, it finds out which CH it belongs to using the strengths of the advertisements.

Data transmission phase

39 | P a g e 8. if the node is a “low energy node”, and the neighbor is ”high energy node”. it sends the message via the neighbor to neighbor CH.

9. if it is a high energy node, it does its duty. (as a normal node or a CH) 10. After considerable rounds of data transmission. go to 1.

5.3. Simulations and analysis

In order to make the validation of our strategy we simulated the algorithm with the network setup given in the last chapter. The network is a random 100 node network in which the nodes are distributed in random locations in a square area of 50mX50m.The radio model used is the same

In order to make the validation of our strategy we simulated the algorithm with the network setup given in the last chapter. The network is a random 100 node network in which the nodes are distributed in random locations in a square area of 50mX50m.The radio model used is the same