Abstract:
A novel nature-inspired algorithm called the Dynamic Virtual Bats Algorithm (DVBA)
is presented in this thesis. DVBA is inspired by a bat’s ability to manipulate frequency
and wavelength of the emitted sound waves when hunting. A role based search has been
developed to improve the diversification and intensification capability of standard Bat
Algorithm (BA). Although DVBA is inspired from bats, like BA, it is conceptually very
different from BA. BA needs a huge number of population size; however, DVBA employs
just two bats to handle the ”exploration and exploitation” conflict which is known as a
real challenge for all optimization algorithms.
Firstly, we study bat’s echolocation ability and next, the most known bat-inspired
algorithm and its modified versions are analyzed. The contributions of this thesis start
reading and imitating bat’s hunting strategies with different perspectives. In the DVBA, there are only two bats: explorer and exploiter bat. While the explorer bat explores the
search space, the exploiter bat makes an intensive search of the local with the highest
probability of locating the desired target. Depending on their location, bats exchange the
roles dynamically.
The performance of the DVBA is extensively evaluated on a suite of 30 bound-constrained
optimization problems from Congress of Evolutionary Computation (CEC) 2014 and
compared with 4 classical optimization algorithm, 4 state-of-the-art modified bat
algorithms, and 5 algorithms from a special session at CEC2014. In addition, DVBA
is tested on supply chain cost problem to see its performance on a complicated real world
problem. The experimental results demonstrated that the proposed DVBA outperform, or
is comparable to, its competitors in terms of the quality of final solution and its convergence
rates.