-This paper presents an improved method for deployment of wireless sensor networks using weighted Artificial Fish Swarm Algorithm (wAFSA). In wireless sensor networks, optimal deployment of sensor nodes is a very critical and important factor in obtaining efficient and reliable quality of service. An improved approach called inertial weight is first introduced into the standard artificial fish swarm algorithm to adaptively select the visual and step sizes of the artificial fishes before the preying, swarming and chasing behaviours of wAFSA were used to randomly and optimally deploy a total of sixty sensor nodes in a network coverage area of 60 square meters. The proposed method was evaluated using some performance metrics such as network coverage and mobile node, network coverage and iteration. The method was also compared with results obtained using standard AFSA which showed that wAFSA performed better and can be used to adequately improve the scalability of wireless sensor networks.
INTRODUCTION
The deployment of sensor nodes is a very important
factor which affects the quality of wireless sensor networks
(WSN). Poor deployment of sensor modes in WSN could
lead to low mode density which in turn causes
communication gap and high node density which causes
message collision and retransmission, signal intrusion and
cramming, huge energy consumption, etc. thereby leading to
challenges in scalability, stability, distributed architecture,
energy consumption, and autonomous operation of the WSN
[1]. The introduction and development of Artificial
Intelligence (AI) has been given significant research
attention in engineering and related disciplines. These
artificial intelligence (heuristic and metaheuristics)
algorithms have shown robustness and strong ability in
solving problems such as WSN deployment and target
tracking as shown in [
The Artificial Fish Swarm Algorithm (AFSA) has been
widely adopted in solving various optimization problems
because of its numerous advantages such as; ease of
implementation, adaptive search ability, high convergence
speed, ability to effectively acquire global solutions while
avoiding local extreme solutions [
II.
The AFSA has been used by several authors in the
deployment and tracking of WSN and a significant
performance when compared with other algorithms have
been observed. The paper presented in [
III.
For the essence of simplicity and easy applicability, this paper considers a wireless sensor network having fixed and mobile sensor nodes. Sixty (60) random nodes were generated in a square coverage area of 60m 60m . This number of fixed and mobile sensor nodes are represented as N and D for easy adaptability into the optimization algorithm that would be used for deployment of these sensor nodes. The significance and important of doing this is to create a matrix search space which is equivalent to the network coverage area for the sensor nodes to be deployed. N represents the population of sensor nodes to be deployed while D represents the dimension of deployment which can be termed as the search space usually in an optimization problem. Therefore, the combination of all nodes in their respective location within the coverage area can be represented as:
N {n1,1, n1,2 ,...ni, j} (1)
Where ni, j represent the ith node in the network, in the jth location of the network. In this case, i has the same size as N and j has the same size as D. where N is the number of sensor nodes and D is the dimension of deployment. Given a random node M with a position (x, y) in the network coverage area and a specific node Q having a position xi , yi within the monitored or specified area, the distance between the nodes can be calculated as [1]: X (M ,Q) Q M (xi x)2 ( yi y)2 (2)
The detection probability of any node Q by another node
M can be calculated based on the probability measurement
model adopted from the work of [
R R X (M ,Q)
s e R Re X (M ,Q) R Re (3) s s
R Re X (M , Q)
s
Where Rs is the perceived radius of various elements in the network, Rs is the uncertainty factor within the measurement range of the nodes for 0 Re R , & s are measured parameters with respect to the physical devices being used and is the input parameters which can be defined as [1]: X (M , P) (Rs R ) e
Therefore, the joint detection probability of multiple sensor nodes when conducting simultaneous measurement simultaneously can be obtained using [1]: P (Q) 1 (1 P (Q, M )) p p wiW (4) (5)
IV.
The Network Coverage Area (NCA) is one of the important indices used to measure the strategy of the wireless sensor network deployment. The network coverage area we considered in this work was the ratio of the whole area that can be covered by the nodes in the total node-aware region and the total area of the monitoring region. Due to the fact that monitoring the environment can be quite complicated, the probability measurement model described in equation (5) was adopted. In order to minimize energy consumption, the following assumptions were made: Network nodes are ideal (i.e. nodes in the network have the same communication radius RC and the same sensor radius RS while RC RS because when the communication radius between the nodes is greater than two times the sensor radius of nodes, then the current networks are connected).
Coverage is considered as a metric for the measurement of quality of service of a sensor network.
At the initial stage of network deployment, all nodes are randomly distributed in a square monitoring area whose length is N, while the coordinate range of the monitoring area is from (0, 0) to (N, D) whereby the D is the dimension of the coverage.
Every node obtains the coordinate position information of itself and its neighbouring nodes.
In order to reduce energy consumption, the movement of nodes is virtual when running the algorithm until after the end of the algorithm, when nodes move to the best location based on the physical location for a single time period.
The model and physical structure of every node is the same and similar.
V.
WEIGHTED ARTIFICIAL FISH SWARM ALGORITHM
The Artificial Fish Swarm Algorithm (AFSA) was
developed based on the intelligent random behaviours of
school of fish which are preying, swarming and chasing
behaviours. It has been established that certain parameters of
AFSA such as visual distance and step size have critical
influence on the performance of the algorithm. For example,
when the visual distance is very large, the algorithm has a
strong ability in obtaining a near optimal global solution and
when the visual distance is small, the algorithm has a strong
local solution searching ability. Furthermore, the bigger the
step size, the faster the speed of convergence of the
algorithm and vice versa [
The ultimate goal of every fish in water is to discover available regions with more food either by vision or by sense through intelligent behaviours such as preying, swarming and chasing. This idea was used in modelling the behaviour of artificial fishes such that the environment where an artificial fish lives is mainly the solution search space of the optimization problem and this defines the state of the fishes. f c max[C(H ')] (8) The basic idea of AFSA is to imitate the preying, swarming and chasing behaviour of fishes. These individual behaviours of AFSA can be represented based on the following rules: Synergic rule: Here, a basic communication is maintained between each individual artificial fish. If an individual fish in the searching swarm receives a call sent by another individual, it moves forward to the calling individual’s position by a step with a certain probability Reconnaissance move rule: If the individual artificial fish does not receive a call, it implements reconnaissance (surveying) according to its own swarm historical experience. If it however finds a better goal, it then moves forward to that position by a step using its respective probability. Stochastic rule: For this rule, if the searching individual fish does not receive a call or does not find a better goal, it moves in steps randomly such that if it then finds a better goal during the step movement, it sends a call to the entire fish swarm.
For any optimization problem, the objective function
under consideration can be said to have a solution space
Dimension-D and the swarm is initialized with P-population
of artificial fishes, such that the state of one artificial fish can
be formulated accordingly. Detail description and relevant
mathematical model for the implementations of wAFSA can
be found in [
The flowchart of the weighted Artificial Fish Swarm Algorithm (wAFSA) as implemented in this paper is shown in Figure 1.
VI.
The problem formulation was done based on the network
area. Here, N.H and H ' represent a set of total nodes and
a set of active nodes respectively. These constitute the total
number of nodes that can be represented as randomly
generated artificial fishes within the solution search space.
The inspection region of H and H’ are correspondingly G
and G’ in which Gi is the inspection region of the ith
artificial fish (node). The network coverage area is
formulated as follows [
H '
N C(H ' ) i1,2,....N
G Gi as:
The quality of the network coverage area is then defined The objective function of the coverage quality which was formulated using equation (7) can be defined as: Apply Inertial
Weight Apply Inertial
Weight
Apply Inertial
Weight
Counter = Counter+1 NO NO
Begin
Initialize Parameters Start counter
Foraging Behavior Succeed? YES Preying Behaviour Succeed? Swarming Behaviour Succeed? Swarming Behaviour Succeed?
YES YES
NO
YES Update Bulletin
Condition Satisfied
YES Output End
NO
NO (6) (7)
Equation (8) represents a maximization objective function which was formulated as a function of network region optimization. The best solution of this objection function returns the optimum value of the network coverage area in which the location of all the sensor nodes are optimized in terms of deployment while consuming minimal energy in the process.
VII. SIMULATION APPROACH
The wAFSA which was previously described was used to optimize the objective function shown in equation (8). Here, mobile sensor nodes are represented as artificial fishes in a cluster form and the cluster head node travels based on the nature of the objective function while following the inertial preying, swarming and chasing movement behaviour. The step by step procedure of the proposed method which is represented in Figure 1 is as follows:
Initialize all the parameters for wAFSA; appears to have little or no significant effect on the performance of the algorithm. Also, for the standard AFSA, the network attains it maximum coverage of about 75.9% when the number of iteration is 20, thereafter, increase in iteration also, appears to have no significant effect on the network. it can be concluded that, even though the wAFSA performed much better in terms of coverage area, the standard AFSA attains it maximum much faster. The improvement of wAFSA over AFSA can be attributed to the adaptive behaviour of the inertial weight which enables to explore larger solution space before it converges. The figure showing the performance evaluation using the coverage area and number of mobile nodes is given in Figure 3.
From Figure 3, it can be observed that, the network coverage increases with increase in number of mobile nodes. For the weighted AFSA, the network coverage attains a maximum of 80.20% at mobile nodes of 50 and then, tends towards stability. Also, for the standard AFSA, a maximum network coverage of about 79.9% was obtained at the same network nodes with similar network behaviours. The insignificant different in the network using both weighted AFSA and standard AFSA indicates the scalability of the network using both approach.
Weighted AFSA Standard AFSA
Initialize the wireless nodes and population of wAFSA to be deployed within the coverage area’s dimension; Initialize iteration start point itr 0 ; Calculate the fitness based on the current locations of the nodes in the coverage area and update the score board with the best individual; Select a new state randomly and evaluate its fitness Evaluate the best fitness based on inertial preying, swarming and chasing behaviour of wAFSA; Evaluate the current fitness of the population (nodes) and compare with previous fitness and then update; If the stopping criteria is met, output the current best fitness, else increase iteration by 1 and go back to step (5) until best solution is obtained.
Display best solution and configuration.
The details of the parameters used for simulation is presented in Table 1. Since the coverage area under consideration is 60 X 60 square metres, 60 wireless sensor nodes were also considered to be deployed randomly and thus, the problem dimension of 60 was used.
Simulations were carried out on MATLAB R2015b and the results are presented based on the performance of the model. The performance evaluations carried out were to check the relationship between network coverage certain parameters such as network node and number of iteration. Readers should note that, the best of every twenty (20) run of the model were presented. The Figure 2 shows the superimposed result obtained for the network coverage against the number of iteration for both the weighted wAFSA and the standard AFSA.
It can be observed from Figure 2 that, the performance of the network coverage increases as the number of iteration is increased from 0 in a step of 5 to 50 iterations. It can be deduced that the wAFSA attained the highest network coverage of about 77.95% which occurs when the number of iterations is 25. At this point, increase in number of iteration
This paper presents an optimized deployment of WSN using a modified artificial fish swarm optimization algorithm (wAFSA). The model was simulated in MATLAB R2015a simulation environment and results show the superiority of the modified algorithm over the standard AFSA. The superiority of the modified algorithm is attributed to the adaptive and dynamic behaviour of the modified algorithm in comparison the original AFSA. In our next research work the performance of the wAFSA will be evaluated on different target tracking and other optimization problems such as data clustering, colour quantization, distributed generation etc.