Repulsion-Propulsion Firefly Algorithm with Fast Convergence to Solve Highly Multi-Modal Problems

Abstract

This paper proposes a modified firefly algorithm (PropFA) to solve highly multimodal problems with fast convergence. In this algorithm every firefly is fitted with a short term memory which facilitates the "Repulsion" force acting on every firefly whereas the introduction of brightness of the best firefly in the movement equation facilitates "Propulsion" force acting on the firefly. Furthermore; no firefly is allowed to move beyond its maximum stride parameter to prevent premature convergence. All other parameters α,β and γ and are designed to be adaptive to suite to almost all multimodal problems without setting them to any a-priori values. Efficiency of PropFA is then verified over six standard test functions. In all of them it is observed that PropFA converges faster than FA and PSO.