Elephant Swarm Optimization In Wireless Sensor Networks

Published Date: 02 Nov 2017

Introduction:

This chapter describes the research work done by author and it has been dedicated to present the real time implementation of the proposed methodology, its mathematical formulation, system integration and validation. In this chapter the author has intended to express his methodological approach for obtaining its proposed cross layered elephant swarm optimization based lifetime maximization of Wireless sensing networks (WSNs).

As discussed earlier, the presented research work has been developed with an expectation of lifetime maximization and its QoS behaviour for overall system optimization. The presented research work has been developed by considering the advantageous features of evolutionary computing. Elephant is a species that exhibits a very sound and effective biological nature and few of the dominating characteristics are like their memory capability, leadership or selection of clan head, group leading and ultimately change of leadership etc. These all types of specific characteristics could play a vital role in network optimization and its lifetime enhancement. Evolutionary computing has ignited a new scenario for research and development in numerous engineering segments. Considering the evolutionary characteristics of elephant swarm behaviour, here in this research work a robust cross layered architecture has been developed. The overall system has been developed while considering the homogenous network conditions and has been simulated with various operating parameters. On the other hand in order to justify the robustness of the proposed work, here in this research the author has developed particle swarm optimization based cross layered architecture as well as most popular protocol called LEACH. The overall system has been developed over SENSORIA, a robust WSNs simulation platform with C# programming language.

Elephant’s swarm behaviour:Elephants are the largest of land mammals living in very advanced social organizations that require good levels of communication between the groups of individuals. This is because they live in a "fluid fission-fusion" society which simply means that their family units are constantly being divided and reunited whilst, at the same time, they are meeting different individuals on a daily basis. This requires an advanced level of communication and recognition to allow individuals to mediate between the complex relationships that they develop with other individuals.

Social organization in elephants is characterized by their closeness and intimacy and is divided into three forms. The most important grouping is the "family unit" which consists of two or more related females and their offspring (See Fig 4.1). Male bulls are not part of this family unit but either aggregate together or solitary.

Elephant herd

Figure 4.1: Elephant exhibiting cluster

A family can range from 2-50 individuals and they interact together in an organized and coordinated manner. Such interactions include teamwork, offspring care, group defense and resource acquisition which all involve decision making, normally made by a powerful "matriarch".

The matriarch is a dominant female leader within the group who is the oldest and thus the wisest and most experienced. She makes the decisions regarding movement, safety and resource acquisition.

"Bond groups", otherwise known as a "kinship groups", are groups of individuals that are closely related genetically. These groups form when bonds in a family are weakened and thus divided by fission. Finally, a "clan" contains approximately 100-250 individuals and are a combination of families and bond groups which share the same home range predominantly during a dry season. They feed together as a big social gathering when resources are scarce and once resources become available, they form a large social organization. During migration, as many as 1000 individuals may aggregate together as a means of protection and dominance during the migration process [147].

Altruism

There is a large degree of altruism and cooperation which is related to kin selection within the family group. This means that one individual will help to increase another individual's lifetime number of offspring and thus help maximize that individual's gene contribution to future generations, but at a cost to their own survival and reproduction. In this case, all young calves are treated equally and are allowed to suckle from other nursing females within the group as well as their mother. Juvenile females act as "aunts" whose role is to make sure the calves behave by preventing them from running ahead of the herd and waking them up after a mid-afternoon nap.

Elephant group banding together for protection

Fig 4.2. Elephant matriarch (right) protecting members of the herd from danger.

Thus, combined parental care is one major factor in elephant social organization. Another important altruistic behaviour observed in elephants is the courageous behaviour of the matriarch, who protects her herd by exposing herself to danger. (See Fig 4.2)

This is an extreme sort of altruism where an individual is not just exposing herself to danger for the sake of another kin, but is increasing the fitness of up to 50 individuals.

Communication

Communication is an essential aspect of social behaviour in elephants and this is achieved by the efficient use of all their senses. This is important as it enables a herd to keep track of their relatives, defending territories and alerting the group of danger, as well as conveying their reproductive state and associating females with their young prior to weaning. Thus, they require a long distance network of communication due to this fluid social system which can convey information about their physical and emotional state as well as transmitting their intentions.

Social Behaviour and Communication in Elephants

Acoustic Communication

Acoustic communication refers to sound production and hearing. Elephants have a wide range of sounds that they can emit all with different intensities and for different purposes such as securing their defense, attracting mates, co-coordinating movement and generally announcing their needs.

These sounds include growls, trumpets, squeals, shrieks and low frequency vocalizations. Growling and rumbling is the most commonly used vocal production in elephants and is used as a means of communication between individuals and families; however, it can also be used in an aggressive tone between females and calves as a disciplinary measure [147].

Angry Elephants Trumpeting & Rumbling

Scientists have now discovered that elephants can produce infrasonic sound from 1- 20Hz that humans cannot hear and these sounds can travel over long distances, as well as seismic signals, which are like mini earthquakes that allow elephants to position each other in relation to their own location [148].

One study suggested that low frequency vocalizations transmitted between females is used as a reproductive strategy by males, specifically in African elephants (Loxonodonta africana). The levels of intensity of these vocalizations vary depending on which reproductive state the females are in. Males use these vocalizations as a strategy to search for herds with high vocal production as this means that they are close to the ovulation period. Males also rely on passive communication between herds by "eavesdropping" to increase his chances of locating a female in the follicular phase and when in close proximity, using chemical and visual cues as more reliable signals [149].

Another study suggests that this "eavesdropping" tactic can also provide an opportunity to recognize signals by co-specifics and this was supported as playback calls made from family and bond groups resulted in a strong positive response of the elephants used in the experiment [150]. Therefore, indicating that elephants have a highly organized network as they can develop and accumulate the knowledge to recognize signals from extensive populations of co-specifics.

Elephant with trunk raised

Fig 4.3. Trunk raised in threatening stance

Visual communication refers to expressions, postures, displays or movement of ears, jaws, trunk and the like, and how elephants are able to use their sense of sight to determine what message the signaler is trying to portray.

Head and trunk postures can be used for different types of communication such as communication between individuals but also between rivals to display threats (See Fig 4.3).

These visual signals can be of either high or low intensity; at low intensity the animal stands tall, but at high intensity i.e. during threat, the animal moves forward towards the enemy lifting its ears and extending its trunk forward. The "forward trunk swish" signal is used towards a smaller rival where the trunk is rolled up and is suddenly lashed forward [147].

To signal dominance, the elephant will appear taller, with its head high above its shoulders and ears spread wide. A subordinate elephant appears the opposite, with its ears kept back and its head lying low [151]. The trunk is thus an important aspect in visual communication.

Chemical Communication

Chemical communication is an energetically efficient process which involves the secretion of chemical signals as long lasting messages. They produce a wide range of odour signals and these odours are carried by secretions from various sources such as skin glands, reproductive tract, urine, faeces and expired air.

Elephant secretion from temporal gland

Fig 4.4. Temporal gland secretion

Secretions are also produced from the temporal gland; a multi-lobed sac which secretes a viscous, strong smelling liquid located between the eye and ear (See Fig 4.4). These secretions are released in great quantities especially when the animal is excited or under stress, suggesting that the gland is under autonomic control [147].

These odour signals are sensed by the receivers' chemo-receptors and are used for many communicative functions such as trail marking, individual recognition and alarm.

Elephants have a highly developed olfactory system which has given them an acute sense of smell that enables the elephant to transfer the chemical into a message.

It has been discovered that elephants can discriminate human friends from unfamiliar humans; African elephants situated in Kenya's Amboseli National Park do not react to the scent or colour of local farmers' garments, but react aggressively and hysterically to the scent and colour of clothing worn by Masai Warriors [152].

This illustrates that elephants can learn by association as different ethnic groups can have different risks to elephants.

Elephants also use chemical communication to detect an individuals' reproductive state. Females prefer males in musth which means they have high testosterone levels and because of this they secrete a fluid from their temporal gland. While in musth they also dribble urine that carries a powerful odour which females can detect.

Musth males are in a good physical condition [153] and are more dominant competitively so females prefer to mate with them rather than those not in the musth phase. This strategy applies to both African and Asian (Elephas maximus) elephants [154].

Tactile Communication

Elephant using trunk to reach vegetation in tall tree

Fig 4.5. Using trunk to pull vegetation off tall tree

Tactile communication is how they use touch to communicate between individuals and they do this primarily with their trunk. Their trunk is used for a variety of different functions such as drinking water, ripping vegetation off tall trees (See Fig 4.5), smelling individuals and so on, but more importantly for their tactile sense.

Elephants use their trunks to explore unfamiliar territories, objects and to exchange touches with unfamiliar individuals passing through the bush. With regards to reproduction, courting elephants communicate with one another by intertwining their trunks.

In addition, feet are also important for their tactile sense as they have soft skin which can sense and hear seismic vibrations through the ground produced by other elephants. The interiors of elephants' feet are filled with Pacinian corpuscles, vibration sensors that are layered and covered in a slimy gel. Vibrations are transmitted through these layers which result in a nerve signal that is sent to the brain.

Their trunks are also used in similar fashion by simply laying them on the ground as they too contain vibration sensors [140].

Memory and Recognition

Elephants living in big social organizations require good memory and recognition capabilities due to the fluid society in which they live. This is because they are constantly separating from the family unit and at the same time, meeting family and non-family members on a daily basis, so need to be able to recognize kin from non-kin; Carol Buckley witnessed two individuals that reunited after 23 years of being apart [140].

Elephant reflected in river

Fig 4.6. Elephant reflection

Elephants have large brains and their cerebral cortex is highly developed which enables them to achieve a greater potential of learning and retaining this information for long periods of time [141].

This gives the elephant a level of intelligence which other land mammals may not have. They are also able to see their reflection in the mirror, a skill that only humans and some primates are able to do. (See Fig 4.6)

One important aspect of recognition and memory in elephants is to avoid inbreeding depression from occurring. Apparently, elephants are able to distinguish paternal kin from non-kin by phenotype matching using olfaction, but the mechanism is still unknown.

The dispersal of one sex in elephants, usually males, is vital to prevent inbreeding depression in addition to their ability to recognize kin and avoid mating with them [142].

Elephants' efficient sense of smell is also used to allow them to make mental maps of their co-specifics' whereabouts and thus track the location of their family members simply by sniffing their urine [143].

One study presented African elephants with urine samples that were either kin or unrelated individuals, situated in locations that were either unexpected or predictable. The elephants' behaviour and reactions to these cues illustrated that they could recognize up to 17 females and 30 family members.

Not only do their senses provide the means for recognition and memory, but their family unit supplies the teaching blocks for young elephants. Calves watch their mothers and older sisters during their days in the bush and learn from them with regards to finding food and water i.e. using their trunk to dig into the ground to reach the water table.

They also make note of how the adults react and behave towards different individuals and they retain this information as a guide during their adulthood. Most actions observed by elephants have been learned during childhood which illustrates their good memories.

Considering the se above mentioned features and characteristics of elephant swarm it can be concluded that elephants are extremely intelligent mammals due to their very advanced sensory systems and social structure. If elephants did not have such powerful sense of smell, touch and the like, their social organization would not be as specialized as it is today with regards to communication and dispersal. It is also their ability to learn from interacting with individuals in their intricate social grouping and to recognize important cues for long periods of time which further adds to their intelligence and complex social network.

Based on the evolutionary characteristics of elephant, here in this research work, the author has proposed a cross layered architecture for optimizing the lifetime of wireless networks (WSNs).

The overall system architecture has been presented as below:

Elephant Swarm Optimization for WSN –A Cross Layer MechanismIn this section of the paper the system modelling adopted to realize the elephant swarm optimization for wireless sensor networks is discussed.

Let us consider a wireless sensor nodes represented by a set which constitute a static network defined as

In the considered network , the wireless communication links that exist between two nodes and , a relatively high transmission power allocation scheme is considered . The high power allocation scheme causes the higher power consumption that ultimately results into numerous interferences situation between other nodes as well as degraded network life time and hence poor efficiency. The communication channel being considered over the links is nothing but Additive White Gaussian Noise () channel having confined noise power level. Here, one more factor called deterministic path loss model has been assumed. If the signal to noise ratio of a communication link is represented by then the maximum data rate supported per unit bandwidth is defined as

Where

This considered model can be realized using modulation schemes like. The constellation size for the isand varies with time over a considered link [23]. The model assumes a scheduling system of communication between the nodes. The model considered assumes that there exists time slots for the medium access control layer () and a unique transmission mode is applicable per slot.

Let us consider that a particular node transmits at a power level then the power consumption of the amplifier is defined as

Where is the efficiency of power amplifier and to achieve the desired signal amplification.

A homogenous sensor network model is considered.

The directed graph that represents the network under consideration, is defined as

Where indicates set of directed links.

Let indicates the incidence matrix of the graph then we can state that

We present an expression

Such that and and have the entries of 0 and 1.

As discussed earlier is the number of time slots in individual frame of the periodic schedule. represents the set of link scheduled. These are allowed to transmit during time slot defined as

and represents the power of transmission and per unit bandwidth rate respectively over link and slot. The vectors of the time slot are and . is the maximum limit of allowable transmission power for the node which belongs to link. The analogous vector is The vectors id defined as

Where is the row of the matrices. Also

The vector is defined as

Where is the row of the matrices.Also

The initial homogenous energy of all the nodes defined as

and the energy

Let represents power consumption of transmitter and represents the power consumption of the receiver and is assumed to be homogenous for all the nodes. The consumed by each node is

Let the sensing events that are induced in the network induce an information generation rate represented as . It can be stated that represents a vector which constitute of.

The data aggregated at the sink is defined as

The link gain matrix of the wireless sensor network considered is defined as

The power from the transmitter of the link to the receiving node on linkis represented as and represents the total noise power over the operational bandwidth.

The represent the network lifetime when a percentage of nodes runs out of energy. This is a common criterion considered by researchers to evaluate their proposed algorithms [16] [17].

The maximum data rate supported for transmission over a particular Linkis defined as

Problem Formulation

A cross layer approach is adopted to enhance the network lifetime of the wireless sensor network. Elephants are social animals and are said to possess strong memory of the events that occur

The problem of optimizing or maximizing the life span of the network can be presented as a function defined as follows

The maximization function can be defined as

For all time slots and, the constituting variables are,,, for a set .

Let us define a variable such that

The elephant swarm optimization is applied to attain minimized function defined as

The minimization function or the elephant swarm optimization objective can be defined as

The model presented here considers based systems the minimization function can be defined as

Whererepresents the link, the number of slots assigned onis. is the set of transmitting links and is the receiving links of the sensor node .The variable is defined as follows

And is defined as

The transmission power thelink represented as is defined as

.

It must be noticed that the power of transmission over a network link is presented as

Cross Layer Optimization to realize Elephant Swarm behaviour

The presented section of this paper elaborates the elephant swarm optimization algorithm for routing, scheduling and advanced radio layer control techniques. The elephant swam optimization is applied taking into account unconstrainedscheduling on the network links.The elephant swarm optimization enables simultaneous scheduling of the sensing data on the interfering wireless communication links in the current consideredscheduling time slot. The elephant swarm optimization iterates to obtain an optimal routing, power consumption and schedule to enhance the considered network lifetime. The elephant model is adopted to solve optimization objective defined in the former section of this paper.

Let us consider a link schedule of data defined as where. The rate of transmission that can be supported over a linkcan be expresses as based on approximations is defined as

If the of a linkis then the minimum transmission rate is defined as follows

.

The elephant swarm optimization results arising based on the above approximations for are said to be a part of the function optimization set.

Let us define a variable .

Then the above equation for the maximum transmission rate optimization can be defined as can be

Based on the above arguments the elephant swarm optimization objective can be expressed as

The above defined elephant swarm optimization is applicable provided

and

In other terms the elephant swarm optimization is applicable if the links have a greater than unity.

The scheduling over all the links is not adopted as the power consumption would exponentially increase. The elephant swarm optimization is applied on all the links scheduled. The computational complexity of optimization under such circumstances can be defined as

From the above equation it is evident that the optimization is computationally heavy and increases exponentially as the links of the sensor nodes increase (i.e. for dense networks) and the slot value increases. The computation complexity of the elephant swarm optimization can be reduced if the number of slots are doubled to. The two fold increase in the number of time slots enables achieving lower power consumption as the sensor nodes have numerous slot options and sleep induction is effective. Furthermore in the case of high sensing activity leading to greater data transmissions, the data to the sink is scheduled using multiple TDMA slots to enable energy conservation and accurate data aggregation.

The elephant swarm optimization model can be summarized in the form of the algorithm given below realized through multiple phases described below.

Phase A:

Initialize the schedule based on the data. The is initialized such that link.the schedule is constructed in a manner such that all the links are provided at least a slot in

Phase B:

In this phase the following equation is solved

If the results obtained on solving are not suitable then the optimization is not possible. If the solutions satisfy the condition then elephant swarm route optimization and radio layer optimizations are carried out to support the required transmission rate.

Phase C:

Evaluate all the links and retain the links if the following equation is satisfied.

This phase eliminates all the links whose is less than unity and retaining the links having an acceptable.

Phase D:

Compute using the following equation

Compute defined as

Using the above definitions we can obtain the new the layer schedule represented as and. If the optimized schedule is equitant to the existing or previous schedule then no further optimization is possible. If the optimized is not similar to the current and previous MAC layer schedule the new schedule is adopted. enables to identify the maximum power utilization link so that it can be scheduled it the additional slots available and thus achieving energy conservation.

Phase E:

In the last phase of the elephant swarm optimization algorithm the optimal solution achieved using a cross layer approach is verified using the following definition

If the solution does not satisfy the above equation then no optimization is possible owing to current network dependent reasons. If optimal solution is obtained and incorporated network performance in terms of data aggregation, improved data rates and higher network lifetimes.

The elephant swarm optimization is realized using a cross layer design approach to enhance network lifetime. The efficiency and the performance measure of this optimization technique is discussed in the subsequent section of this thesis.