Analysis of randomisation methods in swarm intelligence

Nowadays, many stochastic metaheuristics have been developed to solve various optimisation problems. The primary characteristics of these heuristics often involve the use of randomness in their search process. Essentially, randomness is useful when determining the next point in the search space and therefore has a crucial impact when exploring new solutions. In this paper, an extensive comparison is made between various probability distributions that can be used for randomising the swarm intelligence algorithms, e.g., uniform, Gaussian, Lévy flights, chaotic maps, and the random sampling in turbulent fractal cloud. These randomisation methods were incorporated into the bat algorithm that is one of the newest member of this domain. In line with this, various variants of bat algorithms randomised with different randomisation methods have been developed and extensive experiments were conducted on a well-known set of 24 BBOB benchmark functions. In addition, the results of randomised bat algorithms were compared with the results of the other well-known algorithms, including the firefly algorithm, differential evolution and artificial bee colony algorithms. The results of these experiments show that the efficiencies of the distributions used during the tests depend on the problem to be solved as well as on the algorithm used.