TopIntroduction
In recent years, there has been a sudden boom in the application of metaheuristic algorithms to solve optimization problems. The metaheuristic techniques differ from the mathematical programming techniques in that they do not use the gradient of the objective function but heuristic search. The heuristic and meta-heuristic search algorithms are a trial-and-error approach for solving decision-making problems. They employ a rule of thumb with the expectation of reaching the optimum solution, though there is no guarantee for it. A heuristic is a problem-dependent algorithm that exploits problem dependent information to find a sufficiently good solution to a specific problem, while a metaheuristic is a general-purpose algorithm that can be applied to almost any type of optimization problem (Saka et al., 2013; Boussaid et al., 2013).
Some of the prominent metaheuristic techniques are based on Swarm Intelligence (SI). A swarm is a large number of homogenous, unsophisticated agents that interact locally among themselves and their environment, without any central control or management to yield a global behavior. The collective behavior of decentralized and self-organized natural or artificial systems that leads to the solution of complex problems is called swarm intelligence (Kennedy et al., 2001). It is swarm intelligence that enables a colony of ants, for example, to find the shortest route between the nest and the food source or a swarm of bees to spot the locality with maximum amount of nectar.
The two most prominent SI optimization techniques are: Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO). PSO, proposed by Kenedy and Eberhart (1995), and ACO proposed by Dorigo (1992) have become very popular in solving complex and intricate optimization problems in various fields (Yang et al., 2009). Recent developments in ACO and PSO algorithms and their robust industrial applications are found in Chan et al. (2007).
However, despite having several attractive features, it has been observed that these algorithms do not always perform as expected. According to the No Free Lunch Theorem, an algorithm which performs exceptionally well on one problem will perform poorly on another, so much so that the average performance of all metaheuristic algorithms is roughly the same (Wolpert, 1997). PSO, for instance, although performs well on most test and application problems, has the weakness of converging prematurely and falling in local optima. ACO, too, has similar limitations. Several PSO and ACO variants have been developed to overcome these limitations.
The success of the metaheuristics optimization algorithms depends to a large extent on the careful balance between two conflicting goals: exploration (diversification) and exploitation (intensification). While exploration is important to ensure that every part of the solution domain is searched enough to provide a reliable estimate of the global optimum; exploitation, on the other hand, is important to concentrate the search effort around the best solutions found so far by searching their neighborhoods to reach better solutions. The search algorithms achieve these two goals by using local or global search approaches, or an integration of both.
This chapter deals with the hybridization of PSO with ACO, since the two are arguably the most outstanding techniques of the SI paradigm. It identifies four different types of ACO-PSO hybrid strategies: Parallel hybrid (the two algorithms work independently to search for the global optimum), global-local hybrid (one of the algorithms does the global search and the other local search), compound hybrid (the two algorithms themselves are hybridized with other metaheuristic algorithms) and classical hybrid (consisting of the classical discrete ACO and the classical continuous PSO).
This chapter is divided into the following sections: The section on Background describes the background of the swarm intelligence techniques. The section on Swarm Intelligence introduces the ACO and PSO algorithms together with their variants. The section on Hybrid Swarm Intelligence describes the four different ACO-PSO hybrid strategies mentioned above. The section on Applications of ACO-PSO hybrid algorithms explains the hybrid algorithm applications on benchmark problems, in engineering, business and data mining. The section on Future Research Direction indicates the areas of further development of the hybrid algorithms. The chapter ends with a short conclusion and a list of additional readings.