2.3 Optimization Methods for Energy Efficient Routing Protocols
2.3.2 Nature-inspired Optimization Methods
There are many nature-inspired optimization methods proposed in the literature [80, 81]. The optimization methods can be classified into static optimization and dynamic optimization. Static optimization deals with minimizing or maximizing a quantity for a given instant of time, while dynamic optimization refers to the pro- cess of minimizing or maximizing the cost/benefits of some objective function over a period of time [82]. In this study, the problem formulation demands static opti- mization, since optimization process is performed periodically over the operation of the WSNs, at each instance of which the network is temporally static.
(a) Optimization Methods for Cluster-based Network
In cluster-basedWSNs, traditional clustering algorithms likek-means orFCM are used to cluster the network based on the distance measure. In order to achieve
higher efficiency, other parameters like energy need to be considered in the objective function. The more complex problems require optimization methods that can solve the problems in a proficient manner and achieve global or near global optimum.
Evolutionary algorithms eradicate some of the above mentioned difficulties and are quickly replacing the classical methods in solving practical problems [83–
85]. Evolutionary algorithms typically intend to find a good solution to an op- timization problem by trial-and-error in a reasonable amount of computing time.
The most prominent evolutionary algorithm isGAwhich is based on natural genet- ics. GA was developed by John Holland in the 1960s and 1970s [86]. The essence of genetic algorithms involves the encoding of an optimization function into arrays of binary or character strings to represent chromosomes. The population of GA consists of various set of chromosomes. The initial population then evolves by gen- erating new generation individuals via crossover of two randomly selected parent chromosomes, and the mutation of some random bits. Whether a new individual or offspring is selected or not is based on its fitness value, which is determined from the objective function [86, 87]. Later, inspired by the swarm behavior of fish and bird schooling in nature, PSO was developed in 1995 by James Kennedy and Russell C. Eberhart [88]. InPSO, each single solution is represents a particle in the search space. The particles fly through the problem space by following the current optimum particles. During flight, each particle adjusts its position according to its own experience, and the experience of neighboring particles, making use of the best position encountered by itself and its neighbors. The swarm direction of a particle is defined by the set of particles neighboring the particle and its history experience [88, 89]. In WSNs, GA and PSO have been applied to solve several optimization problems. Konstantinidis et al. applied GAto solve a problem of maximizing the
coverage and lifetime of theWSN [91]. The coverage optimization problems of the WSNs are also formulated and solved by usingGA in [92, 93]. The authors in [94]
adoptedPSOto create an optimal power allocation scheme for theWSNto achieve energy savings and reliable communication.
In [31], Latiff et al. applied evolutionary algorithms to obtain the optimiza- tion of CHs selection based on an objective function which contains the residual energy and the relative distance between normal nodes and CHs. The results ob- tained by usingGA and PSO have been compared with and proved to outperform that achieved by some conventional methods.
However, GA inherently suffer from high computational complexity. The most popular versions of GAuse binary encoding and they consume huge memory for achieving high precision. PSO consumes multiple banks of memory to accom- modate the particles and their personal bests. In this study, real-time operation is desired and lighter algorithms are preferred. HSA requires less effort of tuning parameters and is easy to implement. Furthermore, its diversification and intensi- fication are well-controlled [95]. Hence, computational payload may reduced and HSA would be appropriate to support the operation of WSNs.
(b) Optimization Methods for Tree-based Network
Additionally, in thetree-based protocols, sensor nodes are organized into a tree where data aggregation is performed at intermediate nodes, which are located at the junctions of tree branches. The aggregated data packets are later routed to the root node, i. e. theBSs. The tree-based protocols are suitable for applications, that
fire detection, safety monitoring in industrial plants, etc., where the measurement provides the most useful information about the safety conditions. One of the main objectives of the tree-based protocols is to optimize the construction of a data aggregation tree in terms of energy efficiency. This optimal aggregation tree is recognized as NP-Hard [96], which is equivalent to Steiner tree, weighted set cover problem [97].
Greedy Incremental Tree (GIT) is an approximation algorithm for finding optimal aggregation trees based on DD, that is proposed to establish an energy- efficient path in [98]. Krishnamachari et al. compare Shortest Path Tree (SPT), Center at the Nearest Source (CNS) and a modified version of GIT to manifest the advantage of data aggregation and to demonstrate enhanced methods for con- structing the aggregation tree [97].
Misra, R. et al. apply the ant colony algorithm to solve the data aggregation problem [99]. TheACO[100,101] proposed by Dorigo et al., is a well-know swarm intelligent approach which mimics the foraging behaviour of the ant society. These ants deposit pheromone on the ground in order to mark some favorable path that should be followed by the other colony’s member. This approach is utilized in [99]
to determine the optimal paths from different sources to the BSs, data packets are aggregated at the nearest relay nodes to the sources. Liao et al. proposed an improvement of achieving optimal data aggregation tree in WSNs using ACO by extending the search region around the routing paths, and thus increasing the prob- ability of finding aggregation nodes [34]. These works have shown to outperform the conventional methods in terms of energy conservation.