Article Preview
TopIntroduction
Since last decade, nature inspired optimization algorithms has always been a hot topic of research for the optimization community. In the earlier days of computing, many real life complex problems were solved by using some heat and trail methods. Unfortunately, these methods do not work for solving complex real valued problems. So, researchers step forwarded for the development of some competitive optimization algorithms, which are efficient to solve the complex problems. Till date there has been a no. of optimization techniques developed. Most of such techniques are inspired from either nature or any animal or plant species. However, the applications of population based algorithms have captured a special attention.
According to free lunch theorem (Wolpert,1997), there are no such algorithm which is able to solve all type of problem and is relatively better than any of the other competitive algorithm. For any real world single objective problem, the optimal solution is clearly defined, which does not hold for multiobjective problems. The 90’s decade has seen a golden era of using evolutionary techniques such as Genetic Algorithm, Differential Evolution etc. Evolutionary techniques are the stochastic methods, which are based on the simulation of evolution process of nature. As they are able to capture the multiple Pareto-optimal solutions in a distinct simulation run, evolutionary techniques are well suited to solve multiobjective problems as compared to other blind search methods (Fonseca et. al., 1995; Fonseca et. al., 1998; Valenzuela-Rend´on & Uresti-Charre,1997). At some saturation point of view, few researches were against the use of evolutionary techniques due to some facts. As evolutionary algorithms involve a strict fitness computation, which is very complex to calculate, time consuming and affects the computational complexity. Solving certain optimization problems like variant problems is difficult to solve through evolutionary algorithms, due to the poor fitness function and production of bad chromosomes. Another aspect of these algorithms is, if the whole population is in an improved stage, it does not indicate for the improvement of a single solution within that population. Moreover, there is no fixed response time for these algorithms i.e. the gap between the shortest and longest optimization response time is too large. Also, these algorithms do not guarantee for reaching at global optima.